Wednesday Jul 06, 2016

SPECjEnterprise2010: Oracle Server X6-2 Top x86 Result

Two Oracle Server X6-2 systems, using the Intel Xeon E5-2699 v4 processor, produced a world record for x86 systems SPECjEnterprise2010 benchmark result of 27,803.39 SPECjEnterprise2010 EjOPS. One Oracle Server X6-2 system ran the application tier and the second Oracle Server X6-2 system ran the database tier.

  • The Oracle Server X6-2 system demonstrated 44% better performance when compared to the IBM X3650 M5 server result of 19,282.14 SPECjEnterprise2010 EjOPS.

  • The Oracle Server X6-2 system demonstrated 29% better performance when compared to the previous generation Oracle Server X5-2 system result of 21,504.30 SPECjEnterprise2010 EjOPS.

  • This result used Oracle WebLogic Server 12c, Java HotSpot 64-Bit Server 1.8.0_91, Oracle Database 12c, and Oracle Linux.

Performance Landscape

Complete benchmark results are at the SPEC website, SPECjEnterprise2010 Results. The table below shows the top two-chip x86 server results.

SPECjEnterprise2010 Performance Chart
as of 7/6/2016
Submitter EjOPS* Application Server Database Server
Oracle 27,803.39 1 x Oracle Server X6-2
2 x 2.2 GHz Intel Xeon E5-2699 v4
Oracle WebLogic 12c (12.2.1)
1 x Oracle Server X6-2
2 x 2.2 GHz Intel Xeon E5-2699 v4
Oracle Database 12c (12.1.0.2)
Oracle 21,504.30 1 x Oracle Server X5-2
2 x 2.3 GHz Intel Xeon E5-2699 v3
Oracle WebLogic 12c (12.1.3)
1 x Oracle Server X5-2
2 x 2.3 GHz Intel Xeon E5-2699 v3
Oracle Database 12c (12.1.0.2)
IBM 19,282.14 1 x IBM X3650 M5
2 x 2.6 GHz Intel Xeon E5-2697 v3
WebSphere Application Server V8.5
1 x IBM X3850 X6
4 x 2.8 GHz Intel Xeon E7-4890 v2
IBM DB2 10.5

* EjOPS – SPECjEnterprise2010 EjOPS, bigger is better.

Configuration Summary

Application Server:

1 x Oracle Server X6-2
2 x Intel Xeon Processor E5-2699 v4 (2.2 GHz)
256 GB memory
3 x 10 GbE NIC
Oracle Linux 6.7 (kernel-4.1.12-37.2.2.el6uek.x86_64)
Oracle WebLogic Server 12c (12.2.1)
Java HotSpot(TM) 64-Bit Server VM on Linux, version 1.8.0_91 (Java SE 8 Update 91)

Database Server:

1 x Oracle Server X6-2
2 x Intel Xeon Processor E5-2699 v4 (2.2 GHz)
512 GB memory
2 x 10 GbE NIC
1 x 16 Gb FC HBA
2 x Oracle Server X5-2L Storage
Oracle Linux 6.7 (kernel-4.1.12-37.2.2.el6uek.x86_64)
Oracle Database 12c Enterprise Edition Release 12.1.0.2

Benchmark Description

SPECjEnterprise2010 is the third generation of the SPEC organization's J2EE end-to-end industry standard benchmark application. The SPECjEnterprise2010 benchmark has been designed and developed to cover the Java EE 5 specification's significantly expanded and simplified programming model, highlighting the major features used by developers in the industry today. This provides a real world workload driving the Application Server's implementation of the Java EE specification to its maximum potential and allowing maximum stressing of the underlying hardware and software systems.

The workload consists of an end to end web based order processing domain, an RMI and Web Services driven manufacturing domain and a supply chain model utilizing document based Web Services. The application is a collection of Java classes, Java Servlets, Java Server Pages, Enterprise Java Beans, Java Persistence Entities (pojo's) and Message Driven Beans.

The SPECjEnterprise2010 benchmark heavily exercises all parts of the underlying infrastructure that make up the application environment, including hardware, JVM software, database software, JDBC drivers, and the system network.

The primary metric of the SPECjEnterprise2010 benchmark is jEnterprise Operations Per Second ("SPECjEnterprise2010 EjOPS"). This metric is calculated by adding the metrics of the Dealership Management Application in the Dealer Domain and the Manufacturing Application in the Manufacturing Domain. There is no price/performance metric in this benchmark.

Key Points and Best Practices

  • Four Oracle WebLogic server instances were started using numactl binding 2 instances per chip.
  • Four Oracle database listener processes were started, 2 processes bound per processor.
  • Additional tuning information is in the report at spec.org.
  • COD (Cluster on Die) is enabled in the BIOS on the application server.

See Also

Disclosure Statement

SPEC and the benchmark name SPECjEnterprise are registered trademarks of the Standard Performance Evaluation Corporation. Oracle Server X6-2, 27,803.39 SPECjEnterprise2010 EjOPS; Oracle Server X5-2, 21,504.30 SPECjEnterprise2010 EjOPS; IBM System x3650 M5, 19,282.14 SPECjEnterprise2010 EjOPS. Results from www.spec.org as of 7/6/2016.

Wednesday Jun 29, 2016

SPECjEnterprise2010: SPARC S7-2 Secure and Unsecure Results

Oracle's SPARC S7-2 servers produced a SPECjEnterprise2010 benchmark result of 14,400.78 SPECjEnterprise2010 EjOPS using one SPARC S7-2 server for the application tier and one SPARC S7-2 server for the database server.

The SPARC S7-2 servers also obtained a result of 14,121.47 SPECjEnterprise2010 EjOPS using encrypted data. This secured result used Oracle Advanced Security Transparent Data Encryption (TDE) for the application database tablespaces with the AES-128 cipher. The network connection between the application server and the database server was also secured using Oracle's Network Data Encryption with the JDBC driver and RC4-128 cipher.

  • The SPARC S7-2 server, equipped with two SPARC S7 processors, demonstrated 43% better SPECjEnterprise2010 EjOPS/app-server-cores performance compared to the Oracle Server X6-2 system result.

  • The SPARC S7-2 server, equipped with two SPARC S7 processors, demonstrated 50% better SPECjEnterprise2010 EjOPS/app-server-cores performance compared to the Oracle Server X5-2 system result.

  • The SPARC S7-2 server, equipped with two SPARC S7 processors, demonstrated 31% better SPECjEnterprise2010 EjOPS/core performance compared to the 2-chip IBM x3650 server result.

  • The application server used Oracle Fusion Middleware components including the Oracle WebLogic 12.2 application server and Java HotSpot(TM) 64-Bit Server VM on Solaris, version 1.8.0_92. The database server was configured with Oracle Database 12c Release 1.

  • The benchmark performance using the secure SPARC S7-2 server configuration with encryption ran within 2% of the performance of the non-secure SPARC S7-2 server result.

Performance Landscape

Select single application server results. Complete benchmark results are at the SPEC website, SPECjEnterprise2010 Results.

SPECjEnterprise2010 Performance Chart
7/6/2016
Java EE Server DB Server EjOPS app
cores
EjOPS
/appcore
1 x SPARC S7-2
2 x 4.27 GHz SPARC S7
Oracle WebLogic 12c (12.2.1)
1 x SPARC S7-2
2 x 4.27 GHz SPARC S7
Oracle Database 12c (12.1.0.2)
14,400.78 16 900
1 x SPARC S7-2
2 x 4.27 GHz SPARC S7
Oracle WebLogic 12c (12.2.1)
Network Data Encryption for JDBC
1 x SPARC S7-2
2 x 4.27 GHz SPARC S7
Oracle Database 12c (12.1.0.2)
Transparent Data Encryption
14,121.47 16 882
1 x Oracle Server X6-2
2 x 2.2 GHz Intel Xeon E5-2699 v4
Oracle WebLogic 12c (12.2.1)
1 x Oracle Server X6-2
2 x 2.2 GHz Intel Xeon E5-2699 v4
Oracle Database 12c (12.1.0.2)
27,803.39 44 631
1 x Oracle Server X5-2
2 x 2.3 GHz Intel Xeon E5-2699 v3
Oracle WebLogic 12c (12.1.3)
1 x Oracle Server X5-2
2 x 2.3 GHz Intel Xeon E5-2699 v3
Oracle Database 12c (12.1.0.2)
21,504.30 36 597
1 x IBM System x3650 M5
2 x 2.6 GHz Intel Xeon E5-2697 v3
WebSphere Application Server V8.5
1 x IBM System x3850 X6
4 x 2.8 GHz Intel Xeon E7-4890 v2
IBM DB2 10.5 FP5
19,282.14 28 688
1 x IBM Power S824
4 x 3.5 GHz POWER 8
WebSphere Application Server V8.5
1 x IBM Power S824
4 x 3.5 GHz POWER 8
IBM DB2 10.5 FP3
22,543.34 24 939

EjOPS – SPECjEnterprise2010 EjOPS (bigger is better)
app cores – application server cores used
EjOPS/appcore – SPECjEnterprise2010 EjOPS divided by total application server cores (bigger is better)

Configuration Summary

Application Server:

1 x SPARC S7-2 server, with
2 x SPARC S7 processor (4.27 GHz)
512 GB memory (16 x 32 GB)
2 x 600 GB SAS-3 HDD
2 x 400 GB SAS-3 SSD
2 x Sun Dual Port 10 GBase-T Network Adapter
Oracle Solaris 11.3
Oracle WebLogic Server 12c (12.2.1)
Java HotSpot(TM) 64-Bit Server VM on Solaris, version 1.8.0_92

Database Server:

1 x SPARC S7-2 server, with
2 x SPARC S7 processor (4.27 GHz)
512 GB memory (16 x 32 GB)
2 x 600 GB SAS-3 HDD
1 x Sun Dual Port 10 GBase-T Network Adapter
2 x Sun Storage 16 Gbit Fibre Channel Universal HBA
Oracle Solaris 11.3
Oracle Database 12c (12.1.0.2)

Storage Servers:

1 x Oracle Server X6-2L (24 Slot Disk-Cage), with
2 x Intel Xeon Processor E5-2643 v4 (3.4 GHz)
256 GB memory
1 x Sun Storage 16 Gbit Fibre Channel PCI-E HBA dual port
2 x 3.2 TB NVMe SSD
2 x 600 GB SAS-3 HDD
Oracle Solaris 11.3 (11.3.8.0.4)

1 x Oracle Server X6-2L (24 Slot Disk Cage), with
2 x Intel Xeon Processor E5-2643 v4 (3.4 GHz)
256 GB memory
1 x Sun Storage 16 Gb Fibre Channel PCI-E HBA dual port
14 x 600 GB 10k RPM SAS-3 HDD
Oracle Solaris 11.3 (11.3.8.0.4)

1 x Brocade 6510 16 Gb FC switch

Benchmark Description

SPECjEnterprise2010 is the third generation of the SPEC organization's J2EE end-to-end industry standard benchmark application. The SPECjEnterprise2010 benchmark has been designed and developed to cover the Java EE 5 specification's significantly expanded and simplified programming model, highlighting the major features used by developers in the industry today. This provides a real world workload driving the Application Server's implementation of the Java EE specification to its maximum potential and allowing maximum stressing of the underlying hardware and software systems,

  • The web zone, servlets, and web services
  • The EJB zone
  • JPA 1.0 Persistence Model
  • JMS and Message Driven Beans
  • Transaction management
  • Database connectivity
Moreover, SPECjEnterprise2010 also heavily exercises all parts of the underlying infrastructure that make up the application environment, including hardware, JVM software, database software, JDBC drivers, and the system network.

The primary metric of the SPECjEnterprise2010 benchmark is jEnterprise Operations Per Second (SPECjEnterprise2010 EjOPS). The primary metric for the SPECjEnterprise2010 benchmark is calculated by adding the metrics of the Dealership Management Application in the Dealer Domain and the Manufacturing Application in the Manufacturing Domain. There is NO price/performance metric in this benchmark.

See Also

Disclosure Statement

SPEC and the benchmark name SPECjEnterprise are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 7/6/2016. SPARC S7-2, 14,400.78 SPECjEnterprise2010 EjOPS (unsecure); SPARC S7-2, 14,121.47 SPECjEnterprise2010 EjOPS (secure); Oracle Server X6-2, 27,803.39 SPECjEnterprise2010 EjOPS (unsecure); Oracle Server X5-2, 21,504.30 SPECjEnterprise2010 EjOPS (unsecure); IBM Power S824, 22,543.34 SPECjEnterprise2010 EjOPS (unsecure); IBM System x3650 M5, 19,282.14 SPECjEnterprise2010 EjOPS (unsecure).

Real-Time Enterprise: SPARC S7-2 Advantage Per Core Under Load Compared to 2-Chip x86 E5-2699 v3

A goal of the modern business is real-time enterprise where analytics are run simultaneously with transaction processing on the same system to provide the most effective decision making. Oracle Database 12c Enterprise Edition utilizing the Oracle In-Memory option is designed for the same database to be able to perform transactions at the highest performance and to perform analytical calculations that once took days or hours to complete orders of magnitude faster.

Oracle's SPARC S7 processor has deep innovations to take the real-time enterprise to the next level of performance. In this test both OLTP transactions and analytical queries were run in a single database instance using all of the same features of Oracle Database 12c Enterprise Edition including the Oracle In-Memory option in order to compare the advantages of the SPARC S7 processor to the Intel Xeon Processor E5-2699 v3.

The SPARC S7 processor uses the Data Analytics Accelerator (DAX). DAX is not a SIMD instruction, but rather an actual co-processor that offloads in-memory queries which frees the cores up for other processing. The DAX has direct access to the memory bus and can execute scans at near full memory bandwidth. Oracle makes the DAX API available to other applications, so this kind of acceleration is not just available to Oracle Database.

  • Oracle's SPARC S7-2 server ran the in-memory analytics RCDB based queries 2.3x faster per chip under load than a two-chip x86 Intel Xeon Processor E5-2699 v3 server on the 24 stream test. Furthermore, the SPARC S7-2 server ran the in-memory analytics RCDB based queries 5.1x faster per core under load than the same x86 server.

  • The SPARC S7-2 server and the two-chip Intel Xeon Processor E5-2699 v3 server both ran OLTP transactions and the in-memory analytics on the same database instance using Oracle Database 12c Enterprise Edition utilizing the Oracle In-Memory option.

Performance Landscape

The table below compares the SPARC S7-2 server and 2-chip x86 Intel Xeon Processor E5-2699 v3 server while running OLTP and in-memory analytics against tables in the same database instance. The same set of transactions and queries were executed on each system. All of the following results were run as part of this benchmark effort.

Real-Time Enterprise Performance Chart
24 RCDB DSS Streams, 112 OLTP users
Simultaneous Mixed
Workload
SPARC S7-2 2-Chip
x86 E5 v3
SPARC Per Chip
Advantage
SPARC Per Core
Advantage
OLTP Transactions
(Transactions per Second)
195,790 216,302 0.9x 2.0x
Analytic Queries
(Queries per Minute)
107 47 2.3x 5.1x

Configuration Summary

SPARC Server:

1 X SPARC S7-2 server
2 X SPARC S7 processor
512 GB Memory
Oracle Solaris 11.3
Oracle Database 12c Enterprise Edition Release 12.1.0.2.0

x86 Server:

1 X Oracle Server X5-2L
2 X Intel Xeon Processor E5-2699 v3
256 GB Memory
Oracle Linux 6.5 (3.8.13-16.2.1.el6uek.x86_64)
Oracle Database 12c Enterprise Edition Release 12.1.0.2.0

Benchmark Description

The Real-Time Enterprise benchmark simulates the demands of customers who simultaneously run both their OLTP database and the related historical warehouse DSS data that would be based on that OLTP data. It answers the question of how a system will perform when doing data analysis, while at the same time executing real-time on-line transactions.

The OLTP workload simulates an Order Inventory System that exercises both reads and writes with a potentially large number of users that stresses the lock management and connectivity, as well as, database access.

The number of customers, orders and users is fully parametrized. This benchmark is base on 100 GB dataset, 15 million customers, 600 million orders and up to 580 users. The workload consists of a number of transaction types including show-expenses, part-cost, supplier-phone, low-inv, high-inv, update-price, update-phone, update-cost, and new-order.

The real cardinality database (RCDB) schema was created to showcase the potential speedup one may see moving from on disk, row format data warehouse/star schema, to utilizing Oracle Database 12c's In-Memory feature for analytical queries.

The DSS workload consists of, as many as, 2,304 unique queries asking questions such as "In 2014, what was the total revenue of single item orders?" or "In August 2013, how many orders exceeded a total price of $50?" Questions like these can help a company see where to focus for further revenue growth or identify weaknesses in their offerings.

RCDB scale factor 1050 represents a 1.05 TB data warehouse. It is transformed into a star schema of 1.0 TB, and then becomes 110 GB in size when loaded in memory. It consists of 1 fact table, and 4 dimension tables with over 10.5 billion rows. There are 56 columns with most cardinalities varying between 5 and 2,000, a primary key being an example of something outside this range.

The results were obtained running a set of OLTP transactions and analytic queries simultaneously against two schema: a real time online orders system and a related historical orders schema configured as a real cardinality database (RCDB) star schema. The in-memory analytics RCDB queries are executed using the Oracle Database 12c In-Memory columnar feature.

Two reports are generated: one for the OLTP-Perf workload and one for the RCDB DSS workload. For the analytical DSS workload, queries per minute and average query elapsed times are reported. For the OLTP-Perf workload, both transactions-per-second in thousands and OLTP average response times in milliseconds are reported.

Key Points and Best Practices

  • This benchmark utilized the SPARC S7 processor's co-processor DAX for query acceleration.

  • All SPARC S7-2 server results were run with out-of-the-box tuning for Oracle Solaris.

  • All Oracle Server X5-2L system results were run with out-of-the-box tunings for Oracle Linux except for the setting in /etc/sysctl.conf to get large pages for the Oracle Database:

      vm.nr_hugepages=98304

  • To create an in memory area, the following was added to the init.ora:

      inmemory_size = 120g

  • An example of how to set a table to be in memory is below:

      ALTER TABLE CUSTOMER INMEMORY MEMCOMPRESS FOR QUERY HIGH

See Also

Disclosure Statement

Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of June 29, 2016.

The previous information is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle's products remains at the sole discretion of Oracle.

Oracle Advanced Analytics: SPARC T7-4 Beats 4-Chip x86 E7 v3

Oracle's SPARC T7-4 server can deliver up to 8.6x better performance than a four-chip x86 Intel Xeon Processor E7-8895 v3 server running Oracle Advanced Analytics data mining features for scoring/prediction analysis.

The SPARC T7-4 server can deliver up to 3.5x faster performance than a four-chip x86 Intel Xeon Processor E7-8895 v3 server running Oracle Advanced Analytics data mining features for training/learning analysis.

For these scoring/prediction algorithms the SPARC T7-4 server is compared to a four-chip Intel Xeon Processor E7-8895 v3 based server on both a system and per core basis.

  • For the Support Vector Machine algorithm using the Interior Point Method solver (SVM IPM), the SPARC server is 8.6x faster than the x86 server and has a 4.8x advantage per core under load.

  • For the Generalized Linear Model Regression algorithm (GLM Regression), the SPARC server is 6.6x faster than the x86 server and has a 3.7x advantage per core under load.

  • For the Generalized Linear Model Classification algorithm (GLM Classification), the SPARC server is 6.2x faster than the x86 server and has a 3.5x advantage per core under load.

  • For the Support Vector Machine algorithm using the Stochastic Gradient Descent solver (SVM SGD solver), the SPARC is 5.5x faster than the x86 server and has a 3.1x advantage per core under load.

  • For the K-Means algorithm, the SPARC server is 6.3x faster than the x86 server and has a 3.6x advantage per core under load.

  • For the Expectation Maximization algorithm, the SPARC server is 6.1x faster than the x86 server and has a 3.4x advantage per core under load.

For these training/learning algorithms the SPARC T7-4 server is compared to a four-chip Intel Xeon Processor E7-8895 v3 based server on both a system and per core basis.

  • For the Support Vector Machine algorithm using the Interior Point Method solver (SVM IPM), the SPARC server is 3.6x faster than the x86 server and has a 2.0x advantage per core under load.

  • For the Generalized Linear Model Regression algorithm (GLM Regression), the SPARC server is 1.4x faster than the x86 server.

  • For the Generalized Linear Model Classification algorithm (GLM Classification), the SPARC server is 2.1x faster than the x86 server and has a 1.2x advantage per core under load.

  • For the Support Vector Machine algorithm using the Stochastic Gradient Descent solver (SVM SGD solver), the SPARC server is 1.9x faster than the x86 server and has a 1.1x advantage per core under load.

  • For the K-Means algorithm, the SPARC server is 1.4x faster than the x86 server.

  • For the Expectation Maximization algorithm, the SPARC server is 1.7x faster the x86 server.

Oracle Advanced Analytics is an option of Oracle Database.

Training/Learning is the part of Machine Learning (ML) and Statistics that analyzes a sample of data to create a model of what is most interesting for the desired analysis. Typically, this is a compute intensive operation that involves many 64-bit floating-point calculations. The output of the training/learning stage is a model that can analyze huge datasets in a stage called scoring and/or prediction. While training/learning is a very important task, typically most time will be spent in the scoring/prediction state.

Performance Landscape

All of the following results were run as part of this benchmark effort.

Oracle Advanced Analytics Summary
Scoring/Prediction
Method Attributes Run Time (sec)
SPARC
per chip
Advantage
SPARC
per core
Advantage
x86 E7 v3
72 cores total
SPARC T7-4
128 cores total
Supervised
SVM IPM Solver 900 206 24 8.6x 4.8x
GLM Regression 900 166 25 6.6x 3.7x
GLM Classification 900 156 25 6.2x 3.5x
SVM SGD Solver 9000 132 24 5.5x 3.1x
Cluster Model
K-Means 9000 222 35 6.3x 3.6x
Expectation Maximization 9000 243 40 6.1x 3.4x

Oracle Advanced Analytics Summary
Training/Learning
Training/Learning:
Creating Model from data
Attributes Run Time (sec)
SPARC
per chip
Advantage
SPARC
per core
Advantage
x86 E7 v3
72 cores total
SPARC T7-4
128 cores total
Supervised
SVM IPM Solver 900 1442 404 3.6x 2.0x
GLM Classification 900 331 154 2.1x 1.2x
SVM SGD Solver 9000 157 84 1.9x 1.1x
GLM Regression 900 78 55 1.4x 0.8x
Cluster Model
Expectation Maximization 9000 763 455 1.7x 0.9x
K-Means 9000 232 161 1.4x 0.8x

Configuration Summary

SPARC Configuration:

SPARC T7-4
4 x SPARC M7 processors (4.13 GHz, 32 cores per chip)
2 TB memory
Oracle Solaris 11.3
Oracle Database 12c Enterprise Edition (12.2 pre-release)

x86 Configuration:

Oracle Server X5-4
4 x Intel Xeon Processor E7-8895 v3 (2.6 GHz, 18 cores per chip)
512 GB memory
Oracle Linux 6.6
Oracle Database 12c Enterprise Edition (12.2 pre-release)

Shared Storage Configuration:

Oracle Server X5-2L with
2 x Intel Xeon Processor E5-2699 v3
4 x Oracle 1.6 TB NVMe SSD
2 x Sun Storage Dual 16Gb Fibre Channel PCIe Universal HBA

Benchmark Description

The benchmark tests various capabilities of Oracle Advanced Analytics.

Statistical analysis was run on historical aviation data, ontime performance for US flights gathered between 1987 and 2014. The dataset sizes were chosen to avoid I/O. The scoring/prediction was tested on a one billion row dataset; the model was first built on a 159 million row dataset. The training/learning was tested on a 640 million row dataset.

The degree of parallelism was set to get the optimal performance for each test. The run actual times of the analysis calculations are reported.

The following algorithms were tested.

Supervised

  • Support Vector Machine algorithm using the Interior Point Method solver (SVM IPM):
    the tables contained about 900 mining attributes

  • Generalized Linear Model Classification algorithm (GLM Classification):
    the tables contained about 900 mining attributes

  • Support Vector Machine algorithm using the Stochastic Gradient Descent solver (SVM SGD solver):
    the tables contained about 9000 mining attributes

  • Generalized Linear Model Regression algorithm (GLM Regression):
    the tables contained about 900 mining attributes

Cluster Model

  • K-Means algorithm:
    the tables contained about 9000 mining attributes

  • Expectation Maximization algorithm:
    the tables contained about 9000 mining attributes

See Also

Disclosure Statement

Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of June 29, 2016.

The previous information is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle's products remains at the sole discretion of Oracle.

In-Memory Database: SPARC S7-2 Performance

Fast analytics on large databases are critical to transforming key business processes. Oracle's SPARC S7 processors are specifically designed to accelerate in-memory analytics using Oracle Database 12c Enterprise Edition and its In-Memory option. The SPARC S7 processor outperforms an x86 E5-2699 v4 chip by up to 2.8x on analytics queries where all queries were run in-memory. In order to test real world deep analysis on the SPARC S7 processor, a scenario with over 2,300 analytical queries was run against a real cardinality database (RCDB) star schema.

The SPARC S7 processor does this by using its Data Analytics Accelerator (DAX). DAX is not a SIMD instruction, but rather an actual co-processor that offloads in-memory queries to free up the cores for other processing. The DAX has direct access to the memory bus and can execute scans at near full memory bandwidth. This kind of acceleration is not just for the Oracle database. Oracle makes the DAX API available to other applications.

  • Oracle's SPARC S7-2 server delivers up to a 2.8x Query Per Minute speedup over a 2-chip x86 E5-2699 v4 server when executing analytical queries using the In-Memory option of Oracle Database 12c.

  • The SPARC S7-2 server scanned over 36 billion rows per second through the database.

  • Oracle Database In-Memory compresses the on-disk RCDB star schema by about 6x when using the Memcompress For Query High setting (more information following below) and by nearly 10x compared to a standard data warehouse row format version of the same database.

Performance Landscape

All of the following results were run as part of this benchmark effort.

RCDB Performance

RCDB Performance Chart
2,304 Queries
Metric 2-Chip
x86 E5-2699 v4
SPARC S7-2 SPARC
Advantage
Elapsed Seconds 1885 sec 675 sec 2.8x
Queries Per Minute 73 qpm 205 qpm 2.8x

Compression

This performance test was run on a Scale Factor 1750 database, which represents a 1.75 TB row format data warehouse. The database is then transformed into a star schema which ends up around 1.1 TB in size. The star schema is then loaded in memory with a setting of "MEMCOMPRESS FOR QUERY HIGH", which focuses on performance with somewhat more aggressive compression. This memory area is a separate part of the System Global Area (SGA) which is defined by the database initialization parameter "inmemory_size". See below for an example. The LINEORDER fact table, which comprises nearly the entire database size, is listed below in memory with its compression ratio.

Column Name Original Size
(Bytes)
In Memory
Size (Bytes)
Compression
Ratio
LINEORDER 1,101,950,197,760 178,583,568,384 6.2x

Configuration Summary

SPARC Server:

1 x SPARC S7-2 server with
2 x SPARC S7 processors, 4.26 GHz
512 GB memory
Oracle Solaris 11.3
Oracle Database 12c Enterprise Edition Release 12.1.0.2
Oracle Database In-Memory

x86 Server:

1 x Oracle Server X6-2L system with
2 x Intel Xeon Processor E5-2699 v4, 2.2 GHz
512 GB memory
Oracle Linux 7.2 (3.8.13-98.7.1.el7uek.x86_64)
Oracle Database 12c Enterprise Edition Release 12.1.0.2
Oracle Database In-Memory

Benchmark Description

The real cardinality database (RCDB) benchmark was created to showcase the potential speedup one may see moving from on disk, row format data warehouse/Star Schema, to utilizing the Oracle Database 12c In-Memory feature for analytical queries. All tests presented are run in-memory.

The workload consists of 2,304 unique queries asking questions such as "In 2014, what was the total revenue of single item orders?" or "In August 2013, how many orders exceeded a total price of $50?" Questions like these can help a company see where to focus for further revenue growth or identify weaknesses in their offerings.

RCDB scale factor 1750 represents a 1.75 TB data warehouse. It is transformed into a star schema of 1.1 TB and then becomes 179 GB in size when loaded in memory. It consists of 1 fact table and 4 dimension tables with over 10.5 billion rows. There are 56 columns with most cardinalities varying between 5 and 2,000. A primary key is an example of something outside this range.

One problem with many industry standard generated databases is that as they have grown in size, the cardinalities for the generated columns have become exceedingly unrealistic. For instance, one industry standard benchmark uses a schema where at scale factor 1 TB, it calls for the number of parts to be SF * 800,000. A 1 TB database that calls for 800 million unique parts is not very realistic. Therefore RCDB attempts to take some of these unrealistic cardinalities and size them to be more representative of, at least, a section of customer data. Obviously, one cannot encompass every database in one schema. This is just an example.

We carefully scaled each system so that the optimal number of users was run on each system under test so that we did not create artificial bottlenecks. Each user ran an equal number of queries and the same queries were run on each system, allowing for a fair comparison of the results.

Key Points and Best Practices

  • This benchmark utilized the SPARC S7 processor's DAX for query acceleration.

  • All SPARC S7-2 server results were run with out-of-the-box tuning for Oracle Solaris.

  • All Oracle Server X6-2L system results were run with out of the box tunings for Oracle Linux, except for the setting in /etc/sysctl.conf to get large pages for the Oracle Database:

    • vm.nr_hugepages=64520

  • To create an in-memory area, the following was added to the init.ora:

      inmemory_size = 200g

  • An example of how to set a table to be in-memory is below:

      ALTER TABLE CUSTOMER INMEMORY MEMCOMPRESS FOR QUERY HIGH

See Also

Disclosure Statement

Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of June 29, 2016.

The previous information is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle's products remains at the sole discretion of Oracle.

Oracle Communications ASAP – Telco Subscriber Activation: SPARC S7-2L Fastest Recorded Result

Oracle's SPARC S7-2L server delivered fastest recorded results on Oracle Communications ASAP. The SPARC S7-2L server ran Oracle Solaris 11.3 with Oracle Database 12c and Oracle Communications ASAP version 7.3, with Oracle Database 11g Release 2 Client.

  • Running Oracle Communications ASAP, the SPARC S7-2L server delivered a fastest recorded result of 3,292 ASDLs/sec (atomic network activation actions).

  • The SPARC S7-2L server running a single instance of the Oracle Communications ASAP application, with both the application and database tiers consolidated onto a single machine, easily supported the service activation volumes of 3,292 ASDLs/sec which is representative of a typical mobile operator with more than 100 million subscribers.

Performance Landscape

All of the following results were run as part of this benchmark effort.

ASAP 7.2 Test Results – 16 NEP
System ASDLs/sec CPU Usage
SPARC S7-2L 3,292 63%

Configuration Summary

Hardware Configuration:

SPARC S7-2L server
2 x SPARC S7 processors (4.27 GHz)
512 GB memory
Flash Storage

Software Configuration:

Oracle Communications ASAP 7.3 Version B122
Oracle Solaris 11.3
Oracle Database 12c Release 12.1.0.2.0
Oracle WebLogic Server 12.1.3.0.0
Oracle JDK 7 update 95

Benchmark Description

Oracle Communications ASAP provides a convergent service activation platform that automatically activates customer services in a heterogeneous network and IT environment. It supports the activation of consumer and business services in fixed and mobile domains against network and IT applications.

ASAP enables rapid service design and network technology introduction by means of its metadata-driven architecture, design-time configuration environment, and catalog of pre-built activation cartridges to reduce deployment time, cost, and risk. The application has been deployed for mobile (3G, 4G and M2M) services and fixed multi-play (broadband, voice, video, and IT) services in telecommunications, cable and satellite environments as well as for business voice, data, and IT cloud services.

It may be deployed in a fully integrated manner as part of the Oracle Communications Service Fulfillment solution or directly integrated with third- party upstream systems. Market-proven for high-volume performance and scalability, Oracle Communications ASAP is deployed by more than 75 service providers worldwide and activates services for approximately 250 million subscribers globally.

The throughput of ASAP is measured in atomic actions per second (or ASDLs/sec). An atomic action is a single command or operation that can be executed on a network element. Atomic actions are grouped together to form a common service action, where each service action typically relates to an orderable item, such as "GSM voice" or "voice mail" or "GSM data". One or more service actions are invoked by an order management system via an activation work order request.

The workload resembles a typical mobile order to activate a GSM subscriber. A single service action to add a subscriber consists of seven atomic actions where each atomic action executes a command on a network element. Each network element was serviced by a dedicated Network Element Processor (NEP). The ASAP benchmark can vary the number of NEPs, which correlate to the complexity of a Telco operator's environment.

See Also

  • SPARC S7-2L Server
    oracle.com     OTN     Blog
  • Oracle Communications ASAP
    oracle.com
  • Oracle Solaris
    oracle.com     OTN     Blog
  • Oracle Database
    oracle.com     OTN     Blog
  • Oracle Fusion Middleware
    oracle.com     OTN     Blog
  • Disclosure Statement

    Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of June 29, 2016.

    Oracle Berkeley DB: SPARC S7-2 Performance

    Oracle's SPARC S7-2 server shows higher throughput performance per core running a mixed transaction workload using Oracle Berkeley DB on 4 simultaneous 20 GB database instances compared to results on a single processor domain of Oracle's SPARC M7-16 server. Each instance contains a sum of 50 million rows of customer, account and orders data.

    • The SPARC S7-2 server delivered a rate of 142,481 transactions per second on the throughput test.

    Performance Landscape

    All of the following results were run as part of this benchmark effort.

    Mixed Workload - Berkeley DB
    Processor Total Cores Perf (tpS) Perf/Core
    SPARC M7 32 266,450 8,327
    SPARC S7 16 142,481 8,905

    Configuration Summary

    Systems Under Test:

    1 x SPARC S7-2 server with
    2 x SPARC S7 processors, 4.26 GHz, 8 cores per processor
    512 GB memory
    Oracle Solaris 11.3
    Oracle Berkeley DB 6.2

    1 x SPARC M7-16 server using a single processor domain with
    1 x SPARC M7 processor, 4.13 GHz, 32 cores per processor
    512 GB Memory
    Oracle Solaris 11.3
    Oracle Berkeley DB 6.2

    Benchmark Description

    The benchmark consists of workload running against a schema of 6 tables: 4 tables that get updated: account, branch, teller, history, 2 read-only: customer and orders. The workload has a set of 4 transactions:

    1. account update: update account, branch, teller balances.
    2. get-order-customer: random read on order to get customer key, then locate and read customer records.
    3. search-order: Get a range of orders.
    4. search-customer: Get a random customer record.
    Transaction mix: account update is 5%; the other three (read-only) 95%. The benchmark sampling time was 5 minutes and the total throughput was calculated.

    Key Points and Best Practices

    • The default mechanism for implementing the Oracle Berkeley DB cache is memory mapped files. Improved performance is obtained using shared memory. For this demonstration, changes are made in the test programs.

    • Changing from shared memory to ISM requires a simple change to the provided Oracle Berkeley DB source code. To add ISM support, the routine os_map.c was modified

      • original: if((infop->addr = shmat(id,NULL,0)) == (void *)-1)
      • ISM: if((infop->addr = shmat(id,NULL,SHM_SHARE_MMU)) == (void *)-1)

    See Also

    Disclosure Statement

    Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 06/29/2016.

    Tuesday May 31, 2016

    SAP Two-Tier Standard Sales and Distribution SD Benchmark: SPARC M7-8 World Record 8 Processors

    Oracle's SPARC M7-8 server produced a world record result for 8-processors on the SAP two-tier Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement Package 5 for SAP ERP 6.0 (8 chips / 256 cores / 2048 threads).

    • The SPARC M7-8 server achieved 130,000 SAP SD benchmark users running the two-tier SAP Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement Package 5 for SAP ERP 6.0.

    • The SPARC M7-8 server is 1.5x faster per core than x86-based HPE Integrity Superdome X running the two-tier SAP Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement Package 5 for SAP ERP 6.0.

    • The SPARC M7-8 server result was run with Oracle Solaris 11 and used Oracle Database 12c.

    • Previously the SPARC T7-2 server set the 2-chip server world record achieving 30,800 SAP SD benchmark users running the two-tier SAP Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement Package 5 for SAP ERP 6.0.

    Performance Landscape

    SAP-SD 2-tier performance table in decreasing performance order with SAP ERP 6.0 Enhancement Package 5 for SAP ERP 6.0 results (current version of the benchmark as of May 2012).

    SAP SD Two-Tier Benchmark
    System
    Processor
    OS
    Database
    Users Resp Time
    (sec)
    Users/
    core
    Cert#
    SPARC M7-8
    8 x SPARC M7 (8x 32core)
    Oracle Solaris 11
    Oracle Database 12c
    130,000 0.93 508 2016020
    HPE Integrity Superdome X
    16 x Intel E7-8890 v3 (16x 18core)
    Windows Server 2012 R2
    Datacenter Edition
    SQL Server 2014
    100,000 0.99 347 2016002

    Number of cores presented are per chip, to get system totals, multiple by the number of chips.

    Complete benchmark results may be found at the SAP benchmark website http://www.sap.com/benchmark.

    Configuration Summary and Results

    Database/Application Server:

    1 x SPARC M7-8 server with
    8 x SPARC M7 processors (4.13 GHz, total of 8 processors / 256 cores / 2048 threads)
    4 TB memory
    Oracle Solaris 11.3
    Oracle Database 12c

    Database Storage:
    7 x Sun Server X3-2L each with
    2 x Intel Xeon Processors E5-2609 (2.4 GHz)
    16 GB memory
    4 x Sun Flash Accelerator F40 PCIe Card
    12 x 3 TB SAS disks
    Oracle Solaris 11

    REDO log Storage:
    1 x Pillar FS-1 Flash Storage System, with
    2 x FS1-2 Controller (Netra X3-2)
    2 x FS1-2 Pilot (X4-2)
    4 x DE2-24P Disk enclosure
    96 x 300 GB 10000 RPM SAS Disk Drive Assembly

    Certified Results (published by SAP)

    Number of SAP SD benchmark users: 130,000
    Average dialog response time: 0.93 seconds
    Throughput:
      Fully processed order line items per hour: 14,269,670
      Dialog steps per hour: 42,809,000
      SAPS: 713,480
    Average database request time (dialog/update): 0.018 sec / 0.039 sec
    SAP Certification: 2016020

    Benchmark Description

    The SAP Standard Application SD (Sales and Distribution) Benchmark is an ERP business test that is indicative of full business workloads of complete order processing and invoice processing, and demonstrates the ability to run both the application and database software on a single system. The SAP Standard Application SD Benchmark represents the critical tasks performed in real-world ERP business environments.

    SAP is one of the premier world-wide ERP application providers, and maintains a suite of benchmark tests to demonstrate the performance of competitive systems on the various SAP products.

    See Also

    Disclosure Statement

    Two-tier SAP Sales and Distribution (SD) standard application benchmarks, SAP Enhancement Package 5 for SAP ERP 6.0 as of 5/14/16:

    SPARC M7-8 (8 processors, 256 cores, 2048 threads) 130,000 SAP SD users, 8 x 4.13 GHz SPARC M7, 4 TB memory, Oracle Database 12c, Oracle Solaris 11, Cert# 2016020
    SPARC T7-2 (2 processors, 64 cores, 512 threads) 30,800 SAP SD users, 2 x 4.13 GHz SPARC M7, 1 TB memory, Oracle Database 12c, Oracle Solaris 11, Cert# 2015050
    HPE Integrity Superdome X (16 processors, 288 cores, 576 threads) 100,000 SAP SD users, 16 x 2.5 GHz Intel Xeon Processor E7-8890 v3 4096 GB memory, SQL Server 2014, Windows Server 2012 R2 Datacenter Edition, Cert# 2016002

    SAP, R/3, reg TM of SAP AG in Germany and other countries. More info www.sap.com/benchmark

    Thursday Mar 24, 2016

    PeopleSoft Human Capital Management 9.1 FP2: SPARC M7-8 Results Using Oracle Advances Security Transparent Data Encryption

    Using Oracle Advanced Security Transparent Data Encryption (TDE), Oracle's SPARC M7-8 server using Oracle's SPARC M7 processor's software in silicon cryptography instructions produced results on Oracle's PeopleSoft Human Capital Management 9.1 FP2 Benchmark that were nearly identical to results run without TDE (clear-text runs). The benchmark consists of three different components, PeopleSoft HR Self-Service Online, PeopleSoft Payroll Batch, and the combined PeopleSoft HR Self-Service Online and PeopleSoft Payroll Batch. The benchmarks were run on a virtualized two-chip, 1 TB LDom of the SPARC M7-8 server.

    Using TDE enforces data-at-rest encryption in the database layer. Applications and users authenticated to the database continue to have access to application data transparently (no application code or configuration changes are required), while attacks from OS users attempting to read sensitive data from tablespace files and attacks from thieves attempting to read information from acquired disks or backups are denied access to the clear-text data.

    • The PeopleSoft HR online-only and the PeopleSoft HR online combined with PeopleSoft Payroll batch showed similar Search/Save average response times using TDE compared to the corresponding clear-text runs.

    • The PeopleSoft Payroll batch-only run showed only around 4% degradation in batch throughput using TDE compared to the clear-text run.

    • The PeopleSoft HR online combined with PeopleSoft Payroll batch run showed less than 5% degradation in batch throughput (payments per hour) using TDE compared to the clear-text result.

    • On the combined benchmark, the virtualized two-chip LDom of the SPARC M7-8 server with TDE demonstrated around 5 times better Search and around 8 times better Save average response times running nearly double the number of online users for the online component compared to the ten-chip x86 clear-text database solution from Cisco.

    • On the PeopleSoft Payroll batch run and using only a single chip in the virtualized two-chip LDom on the SPARC M7-8 server, the TDE solution demonstrated 1.7 times better batch throughput compared to a four-chip Cisco UCSB460 M4 server with clear-text database.

    • On the PeopleSoft Payroll batch run and using only a single chip in the virtualized two-chip LDom on the SPARC M7-8 server, the TDE solution demonstrated around 2.3 times better batch throughput compared to a nine-chip IBM zEnterprise z196 server (EC 2817-709, 9-way, 8943 MIPS) with clear-text database.

    • On the combined benchmark, the two SPARC M7 processor LDom (in SPARC M7-8) can run the same number of online users with TDE as a dynamic domain (PDom) of eight SPARC M6 processors (in SPARC M6-32) with clear-text database with better online response times, batch elapsed times and batch throughput.

    Performance Landscape

    All results presented are taken from Oracle's PeopleSoft benchmark white papers.

    The first table presents the combined results, running both the PeopleSoft HR Self-Service Online and Payroll Batch tests concurrently.

    PeopleSoft HR Self-Service Online And Payroll Batch Using Oracle Database 11g
    System
    Processors
    Chips
    Used
    Users Search/Save Batch Elapsed
    Time
    Batch Pay/Hr
    SPARC M7-8
    (Secure with TDE)
    SPARC M7
    2 35,000 0.55 sec/0.34 sec 23.72 min 1,265,969
    SPARC M7-8
    (Unsecure)
    SPARC M7
    2 35,000 0.67 sec/0.42 sec 22.71 min 1,322,272
    SPARC M6-32
    (Unsecure)
    SPARC M6
    8 35,000 1.80 sec/1.12 sec 29.2 min 1,029,440
    Cisco 1 x B460 M4, 3 x B200 M3
    (Unsecure)
    Intel E7-4890 v2, Intel E5-2697 v2
    10 18,000 2.70 sec/2.60 sec 21.70 min 1,383,816

    The following results are running only the Peoplesoft HR Self-Service Online test.

    PeopleSoft HR Self-Service Online Using Oracle Database 11g
    System
    Processors
    Chips
    Used
    Users Search/Save
    Avg Response Times
    SPARC M7-8 (Secure with TDE)
    SPARC M7
    2 40,000 0.52 sec/0.31 sec
    SPARC M7-8 (Unsecure)
    SPARC M7
    2 40,000 0.55 sec/0.33 sec
    SPARC M6-32 (Unsecure)
    SPARC M6
    8 40,000 2.73 sec/1.33 sec
    Cisco 1 x B460 M4, 3 x B200 M3 (Unsecure)
    Intel E7-4890 v2, Intel E5-2697 v2
    10 20,000 0.35 sec/0.17 sec

    The following results are running only the Peoplesoft Payroll Batch test. For the SPARC M7-8 server results, only one of the processors was used per LDom. This was accomplished using processor sets to further restrict the test to a single SPARC M7 processor.

    PeopleSoft Payroll Batch Using Oracle Database 11g
    System
    Processors
    Chips
    Used
    Batch Elapsed
    Time
    Batch Pay/Hr
    SPARC M7-8 (Secure with TDE)
    SPARC M7
    1 13.34 min 2,251,034
    SPARC M7-8 (Unsecure)
    SPARC M7
    1 12.85 min 2,336,872
    SPARC M6-32 (Unsecure)
    SPARC M6
    2 18.27 min 1,643,612
    Cisco UCS B460 M4 (Unsecure)
    Intel E7-4890 v2
    4 23.02 min 1,304,655
    IBM z196 (Unsecure)
    zEnterprise (5.2 GHz, 8943 MIPS)
    9 30.50 min 984,551

    Configuration Summary

    System Under Test:

    SPARC M7-8 server with
    8 x SPARC M7 processor (4.13 GHz)
    4 TB memory
    Virtualized as an Oracle VM Server for SPARC (LDom) with
    2 x SPARC M7 processor (4.13 GHz)
    1 TB memory

    Storage Configuration:

    2 x Oracle ZFS Storage ZS3-2 appliance (DB Data) each with
    40 x 300 GB 10K RPM SAS-2 HDD,
    8 x Write Flash Accelerator SSD and
    2 x Read Flash Accelerator SSD 1.6TB SAS
    2 x Oracle Server X5-2L as COMSTAR nodes (DB redo logs & App object cache) each with
    2 x Intel Xeon Processor E5-2630 v3
    32 GB memory
    4 x 1.6 TB NVMe SSD

    Software Configuration:

    Oracle Solaris 11.3
    Oracle Database 11g Release 2 (11.2.0.3.0)
    PeopleSoft Human Capital Management 9.1 FP2
    PeopleSoft PeopleTools 8.52.03
    Oracle Java SE 6u32
    Oracle Tuxedo, Version 10.3.0.0, 64-bit, Patch Level 043
    Oracle WebLogic Server 11g (10.3.5)

    Benchmark Description

    The PeopleSoft Human Capital Management benchmark simulates thousands of online employees, managers and Human Resource administrators executing transactions typical of a Human Resources Self Service application for the Enterprise. Typical transactions are: viewing paychecks, promoting and hiring employees, updating employee profiles, etc. The database tier uses a database instance of about 500 GB in size, containing information for 500,480 employees. The application tier for this test includes web and application server instances, specifically Oracle WebLogic Server 11g, PeopleSoft Human Capital Management 9.1 FP2 and Oracle Java SE 6u32.

    Key Points and Best Practices

    In the HR online along with Payroll batch run, the LDom had one Oracle Solaris Zone of 7 cores containing the Web tier, two Oracle Solaris Zones of 16 cores each containing the Application tier and one Oracle Solaris Zone of 23 cores containing the Database tier. Two cores were dedicated to network and disk interrupt handling. In the HR online only run, the LDom had one Oracle Solaris Zone of 12 cores containing the Web tier, two Oracle Solaris Zones of 18 cores each containing the Application tier and one Oracle Solaris Zone of 14 cores containing the Database tier. 2 cores were dedicated to network and disk interrupt handling. In the Payroll batch only run, the LDom had one Oracle Solaris Zone of 31 cores containing the Database tier. 1 core was dedicated to disk interrupt handling.

    All database data files, recovery files and Oracle Clusterware files for the PeopleSoft test were created with the Oracle Automatic Storage Management (Oracle ASM) volume manager for the added benefit of the ease of management provided by Oracle ASM integrated storage management solution.

    In the application tier on the LDom, 5 PeopleSoft application domains with 350 application servers (70 per domain) were hosted in two separate Oracle Solaris Zones for a total of 10 domains with 700 application server processes.

    All PeopleSoft Application processes and the 32 Web Server JVM instances were executed in the Oracle Solaris FX scheduler class.

    See Also

    Disclosure Statement

    Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of March 24, 2016.

    Thursday Mar 17, 2016

    OLTPbenchmark Workload, Open-Source Benchmark: SPARC T7-1 Performance Beats IBM S824, Beats x86 E5-2699 v3

    OLTPbenchmark is an open-source database benchmarking tool that includes an On-Line Transaction Processing (OLTP) transactional workload derived from the industry standard TPC-C workload.

    Oracle's SPARC T7-1 server demonstrated OLTP performance that is 2.76 times faster per chip than Intel Xeon Processor E5-2699 v3 and 5.47 times faster per chip than an IBM POWER8 (3.5 GHz) processor. This means that a SPARC T7-1 is 1.38 times faster than a 2-chip x86 E5 v3 based server. The SPARC T7-1 server is also 1.37 times faster than an IBM Power System S824 (POWER8) server. On per core performance, the SPARC M7 processor used in the SPARC T7-1 server out performed the IBM POWER8 processor. All of these tests used Oracle Database 12c Release 1 (12.1.0.2) Enterprise Edition for the database.

    Comparing the SPARC T7-1 server to the 2-chip x86 E5 v3 server equipped with two 2.3 GHz Intel Xeon Processor E5-2699 v3, we see the following advantages for the the SPARC T7-1 server.

    • On a per chip basis, the SPARC T7-1 server demonstrated 2.76 times better performance compared to the 2-chip x86 E5 v3 server.

    • At the system level, the SPARC T7-1 server demonstrated 1.38 times better performance compared to the 2-chip x86 E5 v3 server.

    Comparing the SPARC T7-1 server to an IBM Power System S824 server equipped with four 3.5 GHz POWER8 processors (6 cores), we see the following advantages for the the SPARC T7-1 server.

    • On a per chip basis, the SPARC T7-1 server demonstrated nearly 5.47 times better performance compared to an IBM Power System S824 server.

    • On a per core basis the SPARC T7-1 server demonstrated nearly 3% better performance per core compared to an IBM Power System S824 server.

    • At the system level, the SPARC T7-1 server demonstrated nearly 1.37 times better performance compared to the IBM Power System S824 server.

    The OLTPbenchmark transactional workload is based upon the TPC-C benchmark specification. Details of the configuration and parameters used are available in the reports referenced in the See Also section.

    Performance Landscape

    All OLTPbenchmark server results were run as part of this benchmark effort (except as noted). All results are run with Oracle Database 12c Release 1 Enterprise Edition. Results are ordered by TPM/core, highest to lowest.

    OLTPbenchmark Transactional Workload
    Relative Performance to x86 System
    System TPM TPM/chip TPM/core
    SPARC T7-1
    1 x SPARC M7 (32 cores/chip, 32 total)
    1.38x 2.76x 1.55x
    IBM Power System S824
    4 x POWER8 (6 cores/chip, 24 total)
    1.01x 0.50x 1.51x
    Oracle Server X5-2
    2 x Intel E5-2699 v3 (18 cores/chip, 36 total)
    1.00x 1.00x 1.00x

    TPM – OLTPbenchmark transactions per minute

    Results on the IBM Power System S824 were run by Oracle engineers using Oracle Database 12c.

    Configuration Summary

    Systems Under Test:

    SPARC T7-1 server with
    1 x SPARC M7 processor (4.13 GHz)
    512 GB memory
    2 x 600 GB 10K RPM SAS2 HDD
    1 x Sun Dual Port 10 GbE PCIe 2.0 Networking card with Intel 82599 10 GbE Controller
    1 x Sun Storage 16 Gb Fibre Channel Universal HBA
    Oracle Solaris 11.3
    Oracle Database 12c Release 1 (12.1.0.2) Enterprise Edition
    Oracle Grid Infrastructure 12c Release 1 (12.1.0.2)

    Oracle Server X5-2 with
    2 x Intel Xeon processor E5-2699 v3 (2.3 GHz)
    512 GB memory
    2 x 600 GB 10K RPM SAS2 HDD
    1 x Sun Dual Port 10 GbE PCIe 2.0 Networking card with Intel 82599 10 GbE Controller
    1 x Sun Storage 16 Gb Fibre Channel Universal HBA
    Oracle Linux 6.5
    Oracle Database 12c Release 1 (12.1.0.2) Enterprise Edition
    Oracle Grid Infrastructure 12c Release 1 (12.1.0.2)

    IBM Power System S824 with
    4 x POWER8 (3.5 GHz)
    512 GB memory
    4 x 300 GB 15K RPM SAS HDD
    1 x 10 GbE Network Interface
    1 x 16 Gb Fibre Channel HBA
    AIX 7.1 TL3 SP3
    Oracle Database 12c Release 1 (12.1.0.2) Enterprise Edition
    Oracle Grid Infrastructure 12c Release 1 (12.1.0.2)

    Storage Servers:

    1 x Oracle Server X5-2L with
    2 x Intel Xeon Processor E5-2630 v3 (2.4 GHz)
    32 GB memory
    1 x Sun Storage 16 Gb Fibre Channel Universal HBA
    4 x 1.6 TB NVMe SSD
    2 x 600 GB SAS HDD
    Oracle Solaris 11.3

    1 x Oracle Server X5-2L with
    2 x Intel Xeon Processor E5-2630 v3 (2.4 GHz)
    32 GB memory
    1 x Sun Storage 16 Gb Fibre Channel Universal HBA
    14 x 600 GB SAS HDD
    Oracle Solaris 11.3

    Benchmark Description

    The OLTPbenchmark workload as described from the OLTPbenchmark website:

    This is a database performance testing tool that allows you to conduct database workload replay, industry-standard benchmark testing, and scalability testing under various loads, such as scaling a population of users who executes order-entry transactions against a wholesale supplier database.
    OLTPbenchmark supports many databases including Oracle, SQL Server, DB2, TimesTen, MySQL, MariaDB, PostgreSQL, Greenplum, Postgres Plus Advanced Server, Redis and Trafodion SQL on Hadoop.

    Key Points and Best Practices

    • For these tests, an 800 warehouse database was created to compare directly with results posted by Intel.

    • To improve the scalability, the OrderLine table was partitioned and loaded into a separate tablespace using the OLTPbenchmark GUI. The default blocksize was 8K and the OrderLine tablespace blocksize was 16K.

    • To reduce latency of Oracle "cache chains buffers" wait events, the OLTPbenchmark kit was modified by adding partitioning to the NEW_ORDER table as well as the ORDERS_I1 and ORDERS_I2 indexes.

    • To reduce latency of Oracle "library cache: mutex X" wait events, added recommended workarounds from the following Intel blog

    • Refer to the detailed configuration documents in the See Also section below for the list of Oracle parameters.

    See Also

    Disclosure Statement

    Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of March 17, 2016.

    Tuesday Mar 15, 2016

    Oracle Advanced Security – Transparent Data Encryption: Secure Database on SPARC M7 Processor Performance Nearly the Same as Clear

    Oracle's SPARC T7-1 server is faster and more efficient than a two-processor x86 server (Intel Xeon Processor E5-2699 v3) in processing I/O intensive database queries when running the Oracle Advanced Security Transparent Data Encryption (TDE) feature of Oracle Database 12c.

    • The single-processor SPARC T7-1 server is up to 1.4 times faster than the two-processor x86 system for all queries tested, with TDE enabled and without. On a per chip basis, Oracle's SPARC M7 processor is over twice the performance of the Intel Xeon Processor E5-2699 v3 (Haswell).

    • The SPARC T7-1 server is more efficient than the two-processor x86 system for all queries tested, with TDE enabled and without, as measured by CPU utilization. For example, on Query A the CPU utilization nearly doubled on the x86 server (41% on clear to 79% with TDE) while on the same Query A the SPARC T7-1 server CPU utilization 30% on clear to 38% with TDE.

    In a head-to-head comparison of system performance using Oracle's Transparent Data Encryption, the SPARC T7-1 single processor system with one SPARC M7 (4.13 GHz) processor outperforms a two-processor x86 server with Intel Xeon Processor E5-2699 v3 (2.3 GHz) processors. The two systems were configured with the same storage environment, 256 GB of memory, the same version of Oracle Database 12c, and with the same high-level of tunings. All tests run with TDE security used the hardware instructions available on the processors (SPARC or x86).

    Performance Landscape

    In the first table below, results are presented for three different queries and a full table scan. The results labeled "clear" were executed in clear text or without Transparent Data Encryption. The results labeled "TDE" are with AES-128 encryption enabled for all of the data tables used in the tablespace with the default parameter of db_block_checking=false.

    Query Times (seconds – smaller is better)
    System Security Query A Query B Query C Full Table Scan
    SPARC T7-1 clear 64.0 61.0 54.8 52.7
    TDE 65.3 62.8 56.3 53.4
    TDE to Clear ratio 1.02x 1.03x 1.03x 1.01x
    Two x86 E5 v3 clear 69.6 68.6 61.7 54.5
    TDE 89.4 88.7 73.5 58.1
    TDE to Clear ratio 1.3x 1.3x 1.2x 1.1x
    Comparing SPARC and x86 on Query Times
    SPARC advantage – clear 1.09x 1.12x 1.13x 1.03x
    SPARC advantage – TDE 1.37x 1.41x 1.31x 1.09x

    From the table above, the average increase in the query's execution time for the SPARC T7-1 server with TDE enabled is about 2%. The average slow down for the x86 server is about 20%.

    Looking into the utilization of the individual processor's cores, reveals that the single processor SPARC T7-1 server, with 32 cores, has an average core utilization of 36% with TDE enabled. The SPARC T7-1 server still has plenty of cycles and additional processing capability to handle other work. The two-processor x86 E5 v3 server with a total of 36 cores reveals an average core utilization of over 79% with TDE enabled. This means there is little to no room in the processors for handling additional work beyond executing just one of these queries individually without affecting the query's execution time and resources. These results are in the table below.

    Average Core Utilization (smaller is better)
    System Security Query A Query B Query C Full Table Scan
    SPARC T7-1 clear 30% 32% 27% 21%
    TDE 38% 40% 36% 31%
    TDE to Clear ratio 1.3x 1.3x 1.3x 1.5x
    Two x86 E5 v3 clear 41% 40% 38% 41%
    TDE 79% 73% 80% 86%
    TDE to Clear ratio 1.9x 1.8x 2.1x 2.1x
    Comparing SPARC and x86 on Utilization
    SPARC advantage – clear 1.37x 1.25x 1.41x 1.95x
    SPARC advantage – TDE 2.08x 1.83x 2.22x 2.77x

    Configuration Summary

    SPARC Configuration:

    SPARC T7-1 server with
    1 x SPARC M7 processor (4.13 GHz, 32 cores)
    256 GB memory
    Flash storage
    Oracle Solaris 11.3
    Oracle Enterprise Database 12c

    x86 Configuration:

    Oracle Server X5-2L system with
    2 x Intel Xeon Processor E5-2699 v3 (2.3 GHz, 36 total cores)
    256 GB memory
    Flash storage
    Oracle Solaris 11.3
    Oracle Enterprise Database 12c

    Note that the two systems were configured with the same storage environment, the same version of Oracle Database 12c, and with the same high-level of tunings.

    Benchmark Description

    The benchmark executes a set of queries on a table of approximately 1 TB in size. The database contains two copies of the table, one that was built using security and one that does not. The tablespaces used the same layout on the storage and DBMS parameters. Each query is executed individually after a restart of the database and the average of 5 executions of the query is used as the average execution time and the gathering of other system statistics.

    Description of the queries:

    • Query A: Determines how the market share of a given nation within a region has changed over two years for a given part type.
    • Query B: Identifies customers who might have a problem with parts shipped to them.
    • Query C: Determines how much average yearly revenue would be lost if orders were no longer filled for small quantities of certain parts.
    • Full Table Scan: Full table scan of the largest table, over 700 GB of data

    Key Points and Best Practices

    • For each system, the 1 TB of data is spread evenly across the flash storage in 1 MB stripes. This was determined to be the most efficient stripe size for a data warehouse environment with large sequential read operations. With each system having the same amount of memory and database software, the same tuning parameters were used on each system to ensure a fair comparison and that each query induced roughly the same amount of I/O throughput per query.

    • Efficiency was verified by looking at not only the average processor utilization (as measured by Oracle Solaris tool pgstat(1M)), but also by measuring the average processor core utilization at the hardware level.

    See Also

    Disclosure Statement

    Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of March 14, 2016.

    Thursday Nov 19, 2015

    SPECvirt_2013: SPARC T7-2 World Record Performance for Two- and Four-Chip Systems

    Oracle's SPARC T7-2 server delivered a world record SPECvirt_sc2013 result for systems with two to four chips.

    • The SPARC T7-2 server produced a result of 3198 @ 179 VMs SPECvirt_sc2013.

    • The two-chip SPARC T7-2 server beat the best four-chip x86 Intel E7-8890 v3 server (HP ProLiant DL580 Gen9), demonstrating that the SPARC M7 processor is 2.1 times faster than the Intel Xeon Processor E7-8890 v3 (chip-to-chip comparison).

    • The two-chip SPARC T7-2 server beat the best two-chip x86 Intel E5-2699 v3 server results by nearly 2 times (Huawei FusionServer RH2288H V3, HP ProLiant DL360 Gen9).

    • The two-chip SPARC T7-2 server delivered nearly 2.2 times the performance of the four-chip IBM Power System S824 server solution which used 3.5 GHz POWER8 six core chips.

    • The SPARC T7-2 server running Oracle Solaris 11.3 operating system, utilizes embedded virtualization products as the Oracle Solaris 11 zones, which in turn provide a low overhead, flexible, scalable and manageable virtualization environment.

    • The SPARC T7-2 server result used Oracle VM Server for SPARC 3.3 and Oracle Solaris Zones providing a flexible, scalable and manageable virtualization environment.

    Performance Landscape

    Complete benchmark results are at the SPEC website, SPECvirt_sc2013 Results. The following table highlights the leading two-, and four-chip results for the benchmark, bigger is better.

    SPECvirt_sc2013
    Leading Two to Four-Chip Results
    System
    Processor
    Chips Result @ VMs Virtualization Software
    SPARC T7-2
    SPARC M7 (4.13 GHz, 32core)
    2 3198 @ 179 Oracle VM Server for SPARC 3.3
    Oracle Solaris Zones
    HP ProLiant DL580 Gen9
    Intel E7-8890 v3 (2.5 GHz, 18core)
    4 3020 @ 168 Red Hat Enterprise Linux 7.1 KVM
    Lenovo System x3850 X6
    Intel E7-8890 v3 (2.5 GHz, 18core)
    4 2655 @ 147 Red Hat Enterprise Linux 6.6 KVM
    Huawei FusionServer RH2288H V3
    Intel E5-2699 v3 (2.3 GHz, 18core)
    2 1616 @ 95 Huawei FusionSphere V1R5C10
    HP ProLiant DL360 Gen9
    Intel E5-2699 v3 (2.3 GHz, 18core)
    2 1614 @ 95 Red Hat Enterprise Linux 7.1 KVM
    IBM Power S824
    POWER8 (3.5 GHz, 6core)
    4 1370 @ 79 PowerVM Enterprise Edition 2.2.3

    Configuration Summary

    System Under Test Highlights:

    Hardware:
    1 x SPARC T7-2 server, with
    2 x 4.13 GHz SPARC M7
    1 TB memory
    2 Sun Dual Port 10GBase-T Adapter
    2 Sun Storage Dual 16 Gb Fibre Channel PCIe Universal HBA

    Software:
    Oracle Solaris 11.3
    Oracle VM Server for SPARC 3.3 (LDom)
    Oracle Solaris Zones
    Oracle iPlanet Web Server 7.0.20
    Oracle PHP 5.3.29
    Dovecot v2.2.18
    Oracle WebLogic Server Standard Edition Release 10.3.6
    Oracle Database 12c Enterprise Edition (12.1.0.2.0)
    Java HotSpot(TM) 64-Bit Server VM on Solaris, version 1.7.0_85-b15

    Storage:
    3 x Oracle Server X5-2L, with
    2 x Intel Xeon Processor E5-2630 v3 8-core 2.4 GHz
    32 GB memory
    4 x Oracle Flash Accelerator F160 PCIe Card
    Oracle Solaris 11.3

    1 x Oracle Server X5-2L, with
    2 x Intel Xeon Processor E5-2630 v3 8-core 2.4 GHz
    32 GB memory
    4 x Oracle Flash Accelerator F160 PCIe Card
    4x 400 GB SSDs
    Oracle Solaris 11.3

    Benchmark Description

    SPECvirt_sc2013 is SPEC's updated benchmark addressing performance evaluation of datacenter servers used in virtualized server consolidation. SPECvirt_sc2013 measures the end-to-end performance of all system components including the hardware, virtualization platform, and the virtualized guest operating system and application software. It utilizes several SPEC workloads representing applications that are common targets of virtualization and server consolidation. The workloads were made to match a typical server consolidation scenario of CPU resource requirements, memory, disk I/O, and network utilization for each workload. These workloads are modified versions of SPECweb2005, SPECjAppServer2004, SPECmail2008, and SPEC CPU2006. The client-side SPECvirt_sc2013 harness controls the workloads. Scaling is achieved by running additional sets of virtual machines, called "tiles", until overall throughput reaches a peak.

    Key Points and Best Practices

    • The SPARC T7-2 server running the Oracle Solaris 11.3, utilizes embedded virtualization products as the Oracle VM Server for SPARC and Oracle Solaris Zones, which provide a low overhead, flexible, scalable and manageable virtualization environment.

    • In order to provide a high level of data integrity and availability, all the benchmark data sets are stored on mirrored (RAID1) storage

    • Using Oracle VM Server for SPARC to bind the SPARC M7 processor with its local memory optimized the memory use in this virtual environment.

    • This improved result used a fractional tile to fully saturate the system.

    See Also

    Disclosure Statement

    SPEC and the benchmark name SPECvirt_sc are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 11/19/2015. SPARC T7-2, SPECvirt_sc2013 3198@179 VMs; HP ProLiant DL580 Gen9, SPECvirt_sc2013 3020@168 VMs; Lenovo x3850 X6; SPECvirt_sc2013 2655@147 VMs; Huawei FusionServer RH2288H V3, SPECvirt_sc2013 1616@95 VMs; HP ProLiant DL360 Gen9, SPECvirt_sc2013 1614@95 VMs; IBM Power S824, SPECvirt_sc2013 1370@79 VMs.

    Monday Oct 26, 2015

    Real-Time Enterprise: SPARC T7-1 Faster Than x86 E5 v3

    A goal of the modern business is real-time enterprise where analytics are run simultaneously with transaction processing on the same system to provide the most effective decision making. Oracle Database 12c Enterprise Edition utilizing the In-Memory option is designed to have the same database able to perform transactions at the highest performance and to transform analytical calculations that once took days or hours to complete orders of magnitude faster.

    Oracle's SPARC M7 processor has deep innovations to take the real-time enterprise to the next level of performance. In this test both OLTP transactions and analytical queries were run in a single database instance using all of the same features of Oracle Database 12c Enterprise Edition utilizing the In-Memory option in order to compare the advantages of the SPARC M7 processor compared to a generic x86 processor. On both systems the OLTP and analytical queries both took about half of the processing load of the server.

    In this test Oracle's SPARC T7-1 server is compared to a two-chip x86 E5 v3 based server. On analytical queries the SPARC M7 processor is 8.2x faster than the x86 E5 v3 processor. Simultaneously on OLTP transactions the SPARC M7 processor is 2.9x faster than the x86 E5 v3 processor. In addition, the SPARC T7-1 server had better OLTP transactional response time than the x86 E5 v3 server.

    The SPARC M7 processor does this by using the Data Accelerator co-processor (DAX). DAX is not a SIMD instruction set, but rather an actual co-processor that offloads in-memory queries which frees the cores up for other processing. The DAX has direct access to the memory bus and can execute scans at near full memory bandwidth. Oracle makes the DAX API available to other applications, so this kind of acceleration is not just to the Oracle database, it is open.

    The results below were obtained running a set of OLTP transactions and analytic queries simultaneously against two schema: a real time online orders system and a related historical orders schema configured as a real cardinality database (RCDB) star schema. The in-memory analytics RCDB queries are executed using the Oracle Database 12c In-Memory columnar feature.

    • The SPARC T7-1 server and the x86 E5 v3 server both ran OLTP transactions and the in-memory analytics on the same database instance using Oracle Database 12c Enterprise Edition utilizing the In-Memory option.

    • The SPARC T7-1 server ran the in-memory analytics RCDB based queries 8.2x faster per chip than a two-chip x86 E5 v3 server on the 48 stream test.

    • The SPARC T7-1 server delivers 2.9x higher OLTP transaction throughput results per chip than a two-chip x86 E5 v3 server on the 48 stream test.

    Performance Landscape

    The table below compares the SPARC T7-1 server and 2-chip x86 E5 v3 server while running OLTP and in-memory analytics against tables in the same database instance. The same set of transactions and queries were executed on each system.

    Real-Time Enterprise Performance Chart
    48 RCDB DSS Streams, 224 OLTP users
    System OLTP Transactions Analytic Queries
    Trans Per
    Second
    Per Chip
    Advantage
    Average
    Response Time
    Queries Per
    Minute
    Per Chip
    Advantage
    SPARC T7-1
    1 x SPARC M7 (32core)
    338 K 2.9x 11 (msec) 267 8.2x
    x86 E5 v3 server
    2 x Intel E5-2699 v3 (2x 18core)
    236 K 1.0 12 (msec) 65 1.0

    The number of cores listed is per chip.
    The Per Chip Advantage it computed by normalizing to a single chip's performance

    Configuration Summary

    SPARC Server:

    1 X SPARC T7-1 server
    1 X SPARC M7 processor
    256 GB Memory
    Oracle Solaris 11.3
    Oracle Database 12c Enterprise Edition Release 12.1.0.2.10

    x86 Server:

    1 X Oracle Server X5-2L
    2 X Intel Xeon Processor E5-2699 v3
    256 GB Memory
    Oracle Linux 6 Update 5 (3.8.13-16.2.1.el6uek.x86_64)
    Oracle Database 12c Enterprise Edition Release 12.1.0.2.10

    Benchmark Description

    The Real-Time Enterprise benchmark simulates the demands of customers who want to simultaneously run both their OLTP database and the related historical warehouse DSS data that would be based on that OLTP data. It answers the question of how a system will perform when doing data analysis while at the same time executing real-time on-line transactions.

    The OLTP workload simulates an Order Inventory System that exercises both reads and writes with a potentially large number of users that stresses the lock management and connectivity, as well as, database access.

    The number of customers, orders and users is fully parametrized. This benchmark is base on 100 GB dataset, 15 million customers, 600 million orders and up to 580 users. The workload consists of a number of transaction types including show-expenses, part-cost, supplier-phone, low-inv, high-inv, update-price, update-phone, update-cost, and new-order.

    The real cardinality database (RCDB) schema was created to showcase the potential speedup one may see moving from on disk, row format data warehouse/Star Schema, to utilizing Oracle Database 12c's In-Memory feature for analytical queries.

    The workload consists of as many as 2,304 unique queries asking questions such as "In 2014, what was the total revenue of single item orders", or "In August 2013, how many orders exceeded a total price of $50". Questions like these can help a company see where to focus for further revenue growth or identify weaknesses in their offerings.

    RCDB scale factor 1050 represents a 1.05 TB data warehouse. It is transformed into a star schema of 1.0 TB, and then becomes 110 GB in size when loaded in memory. It consists of 1 fact table, and 4 dimension tables with over 10.5 billion rows. There are 56 columns with most cardinalities varying between 5 and 2,000, a primary key being an example of something outside this range.

    Two reports are generated: one for the OLTP-Perf workload and one for the RCDB DSS workload. For the analytical DSS workload, queries per minute and average query elapsed times are reported. For the OLTP-Perf workload, both transactions-per-seconds in thousands and OLTP average response times in milliseconds are reported.

    Key Points and Best Practices

    • This benchmark utilized the SPARC M7 processor's co-processor DAX for query acceleration.

    • All SPARC T7-1 server results were run with out-of-the-box tuning for Oracle Solaris.

    • All Oracle Server X5-2L system results were run with out of the box tunings for Oracle Linux except for the setting in /etc/sysctl.conf to get large pages for the Oracle Database:

      • vm.nr_hugepages=98304

    • To create an in memory area, the following was added to the init.ora:

        inmemory_size = 120g

    • An example of how to set a table to be in memory is below:

        ALTER TABLE CUSTOMER INMEMORY MEMCOMPRESS FOR QUERY HIGH

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 25 October 2015.

    In-Memory Database: SPARC T7-1 Faster Than x86 E5 v3

    Fast analytics on large databases are critical to transforming key business processes. Oracle's SPARC M7 processors are specifically designed to accelerate in-memory analytics using Oracle Database 12c Enterprise Edition utilizing the In-Memory option. The SPARC M7 processor outperforms an x86 E5 v3 chip by up to 10.8x on analytics queries. In order to test real world deep analysis on the SPARC M7 processor a scenario with over 2,300 analytical queries was run against a real cardinality database (RCDB) star schema. This benchmark was audited by Enterprise Strategy Group (ESG). ESG is an IT research, analyst, strategy, and validation firm focused on the global IT community.

    The SPARC M7 processor does this by using Data Accelerator co-processor (DAX). DAX is not a SIMD instruction but rather an actual co-processor that offloads in-memory queries which frees the cores up for other processing. The DAX has direct access to the memory bus and can execute scans at near full memory bandwidth. Oracle makes the DAX API available to other applications, so this kind of acceleration not just for the Oracle database, it is open.

    • The SPARC M7 processor delivers up to a 10.8x Query Per Minute speedup per chip over the Intel Xeon Processor E5-2699 v3 when executing analytical queries using the In-Memory option of Oracle Database 12c.

    • Oracle's SPARC T7-1 server delivers up to a 5.4x Query Per Minute speedup over the 2-chip x86 E5 v3 server when executing analytical queries using the In-Memory option of Oracle Database 12c.

    • The SPARC T7-1 server delivers over 143 GB/sec of memory bandwidth which is up to 7x more than the 2-chip x86 E5 v3 server when the Oracle Database 12c is executing the same analytical queries against the RCDB.

    • The SPARC T7-1 server scanned over 48 billion rows per second through the database.

    • The SPARC T7-1 server compresses the on-disk RCDB star schema by around 6x when using the Memcompress For Query High setting (more information following below) and by nearly 10x compared to a standard data warehouse row format version of the same database.

    Performance Landscape

    The table below compares the SPARC T7-1 server and 2-chip x86 E5 v3 server. The x86 E5 v3 server single chip compares are from actual measurements against a single chip configuration.

    The number of cores is per chip, multiply by number of chips to get system total.

    RCDB Performance Chart
    2,304 Queries
    System Elapsed
    Seconds
    Queries Per
    Minute
    System
    Adv
    Chip
    Adv
    DB Memory
    Bandwidth
    SPARC T7-1
    1 x SPARC M7 (32core)
    381 363 5.4x 10.8x 143 GB/sec
    x86 E5 v3 server
    2 x Intel E5-2699 v3 (2x 18core)
    2059 67 1.0x 2.0x 20 GB/sec
    x86 E5 v3 server
    1 x Intel E5-2699 v3 (18core)
    4096 34 0.5x 1.0x 10 GB/sec

    Fused Decompress + Scan

    The In-Memory feature of Oracle Database 12c puts tables in columnar format. There are different levels of compression that can be applied. One of these is Oracle Zip (OZIP) which is used with the "MEMCOMPRESS FOR QUERY HIGH" setting. Typically when compression is applied to data, in order to operate on it, the data must be:

      (1) Decompressed
      (2) Written back to memory in uncompressed form
      (3) Scanned and the results returned.

    When OZIP is applied to the data inside of an In-Memory Columnar Unit (or IMCU, an N sized chunk of rows), the DAX is able to take this data in its compressed format and operate (scan) directly upon it, returning results in a single step. This not only saves on compute power by not having the CPU do the decompression step, but also on memory bandwidth as the uncompressed data is not put back into memory. Only the results are returned. To illustrate this, a microbenchmark was used which measured the amount of rows that could be scanned per second.

    SAE hpk-uperf

    Compression

    This performance test was run on a Scale Factor 1750 database, which represents a 1.75 TB row format data warehouse. The database is then transformed into a star schema which ends up around 1.1 TB in size. The star schema is then loaded in memory with a setting of "MEMCOMPRESS FOR QUERY HIGH", which focuses on performance with somewhat more aggressive compression. This memory area is a separate part of the System Global Area (SGA) which is defined by the database initialization parameter "inmemory_size". See below for an example. Here is a breakdown of each table in memory with compression ratios.

    Column Name Original Size
    (Bytes)
    In Memory
    Size (Bytes)
    Compression
    Ratio
    LINEORDER 1,103,524,528,128 178,586,451,968 6.2x
    DATE 11,534,336 1,179,648 9.8x
    PART 11,534,336 1,179,648 9.8x
    SUPPLIER 11,534,336 1,179,648 9.8x
    CUSTOMER 11,534,336 1,179,648 9.8x

    Configuration Summary

    SPARC Server:

    1 X SPARC T7-1 server
    1 X SPARC M7 processor
    512 GB memory
    Oracle Solaris 11.3
    Oracle Database 12c Enterprise Edition Release 12.1.0.2.13

    x86 Server:

    1 X Oracle Server X5-2L
    2 X Intel Xeon Processor E5-2699 v3
    512 GB memory
    Oracle Linux 6 Update 5 (3.8.13-16.2.1.el6uek.x86_64)
    Oracle Database 12c Enterprise Edition Release 12.1.0.2.13

    Benchmark Description

    The real cardinality database (RCDB) benchmark was created to showcase the potential speedup one may see moving from on disk, row format data warehouse/Star Schema, to utilizing Oracle Database 12c's In-Memory feature for analytical queries.

    The workload consists of 2,304 unique queries asking questions such as "In 2014, what was the total revenue of single item orders", or "In August 2013, how many orders exceeded a total price of $50". Questions like these can help a company see where to focus for further revenue growth or identify weaknesses in their offerings.

    RCDB scale factor 1750 represents a 1.75 TB data warehouse. It is transformed into a star schema of 1.1 TB, and then becomes 179 GB in size when loaded in memory. It consists of 1 fact table, and 4 dimension tables with over 10.5 billion rows. There are 56 columns with most cardinalities varying between 5 and 2,000, a primary key being an example of something outside this range.

    One problem with many industry standard generated databases is that as they have grown in size the cardinalities for the generated columns have become exceedingly unrealistic. For instance one industry standard benchmark uses a schema where at scale factor 1 TB it calls for the number of parts to be SF * 800,000. A 1 TB database that calls for 800 million unique parts is not very realistic. Therefore RCDB attempts to take some of these unrealistic cardinalities and size them to be more representative of at least a section of customer data. Obviously one cannot encompass every database in one schema, this is just an example.

    We carefully scaled each system so that the optimal number of users was run on each system under test so that we did not create artificial bottlenecks. Each user ran an equal number of queries and the same queries were run on each system, allowing for a fair comparison of the results.

    Key Points and Best Practices

    • This benchmark utilized the SPARC M7 processor's co-processor DAX for query acceleration.

    • All SPARC T7-1 server results were run with out of the box tuning for Oracle Solaris.

    • All Oracle Server X5-2L system results were run with out of the box tunings for Oracle Linux except for the setting in /etc/sysctl.conf to get large pages for the Oracle Database:

      • vm.nr_hugepages=64520

    • To create an in memory area, the following was added to the init.ora:

        inmemory_size = 200g

    • An example of how to set a table to be in memory is below:

        ALTER TABLE CUSTOMER INMEMORY MEMCOMPRESS FOR QUERY HIGH

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 10/25/2015.

    In-Memory Aggregation: SPARC T7-2 Beats 4-Chip x86 E7 v2

    Oracle's SPARC T7-2 server demonstrates better performance both in throughput and number of users compared to a four-chip x86 E7 v2 sever. The workload consists of a realistic set of business intelligence (BI) queries in a multi-user environment against a 500 million row fact table using Oracle Database 12c Enterprise Edition utilizing the In-Memory option.

    • The SPARC M7 chip delivers 2.3 times more query throughput per hour compared to an x86 E7 v2 chip.

    • The two-chip SPARC T7-2 server delivered 13% more query throughput per hour compared to a four-chip x86 E7 v2 server.

    • The two-chip SPARC T7-2 server supported over 10% more users than a four-chip x86 E7 v2 server.

    • Both the SPARC server and x86 server ran with just under 5 second average response time.

    Performance Landscape

    The results below were run as part of this benchmark. All results use 500,000,000 fact table rows and had average cpu utilization of 100%.

    In-Memory Aggregation
    500 Million Row Fact Table
    System Users Queries
    per Hour
    Queries per Hour
    per Chip
    Average
    Response Time
    SPARC T7-2
    2 x SPARC M7 (32core)
    190 127,540 63,770 4.99 (sec)
    x86 E7 v2
    4 x E7-8895 v2 (4x 15core)
    170 112,470 28,118 4.92 (sec)

    The number of cores are listed per chip.

    Configuration Summary

    SPARC Configuration:

    SPARC T7-2
    2 x 4.13 GHz SPARC M7 processors
    1 TB memory (32 x 32 GB)
    Oracle Solaris 11.3
    Oracle Database 12c Enterprise /Edition (12.1.0.2.0)

    x86 Configuration:

    Sun Server X4-4
    4 x Intel Xeon Processor E7-8895 v2 processors
    1 TB memory (64 x 16 GB)
    Oracle Linux Server 6.5 (kernel 2.6.32-431.el6.x86_64)
    Oracle Database 12c Enterprise /Edition (12.1.0.2.0)

    Benchmark Description

    The benchmark is designed to highlight the efficacy of the Oracle Database 12c In-Memory Aggregation facility (join and aggregation optimizations) together with the fast scan and filtering capability of Oracle's in-memory column store facility.

    The benchmark runs analytic queries such as those seen in typical customer business intelligence (BI) applications. These are done in the context of a star schema database. The key metrics are query throughput, number of users and average response times

    The implementation of the workload used to achieve the results is based on a schema consisting of 9 dimension tables together with a 500 million row fact table.

    The query workload consists of randomly generated star-style queries simulating a collection of ad-hoc business intelligence users. Up to 300 concurrent users have been run, with each user running approximately 500 queries. The implementation includes a relatively small materialized view, which contains some precomputed data. The creation of the materialized view takes only a few minutes.

    Key Points and Best Practices

    The reported results were obtained by using the following settings on both systems except where otherwise noted:

    1. starting with a completely cold shared pool
    2. without making use of the result cache
    3. without using dynamic sampling or adaptive query optimization
    4. running all queries in parallel, where
      parallel_max_servers = 1600 (on the SPARC T7-2) or
      parallel_max_servers = 240 (on the Sun Server X4-4)
      each query hinted with PARALLEL(4)
      parallel_degree_policy = limited
    5. having appropriate queries rewritten to the materialized view, MV3, defined as
      SELECT
      /*+ append vector_transform */
      d1.calendar_year_name, d1.calendar_quarter_name, d2.all_products_name,
      d2.department_name, d2.category_name, d2.type_name, d3.all_customers_name,
      d3.region_name, d3.country_name, d3.state_province_name, d4.all_channels_name,
      d4.class_name, d4.channel_name, d5.all_ages_name, d5.age_name, d6.all_sizes_name,
      d6.household_size_name, d7.all_years_name, d7.years_customer_name, d8.all_incomes_name,
      d8.income_name, d9.all_status_name, d9.marital_status_name,
      SUM(f.sales) AS sales,
      SUM(f.units) AS units,
      SUM(f.measure_3) AS measure_3,
      SUM(f.measure_4) AS measure_4,
      SUM(f.measure_5) AS measure_5,
      SUM(f.measure_6) AS measure_6,
      SUM(f.measure_7) AS measure_7,
      SUM(f.measure_8) AS measure_8,
      SUM(f.measure_9) AS measure_9,
      SUM(f.measure_10) AS measure_10
      FROM time_dim d1, product_dim d2, customer_dim_500M_10 d3, channel_dim d4, age_dim d5,
      household_size_dim d6, years_customer_dim d7, income_dim d8, marital_status_dim d9,
      units_fact_500M_10 f
      WHERE d1.day_id = f.day_id AND
      d2.item_id = f.item_id AND
      d3.customer_id = f.customer_id AND
      d4.channel_id = f.channel_id AND
      d5.age_id = f.age_id AND
      d6.household_size_id = f.household_size_id AND
      d7.years_customer_id = f.years_customer_id AND
      d8.income_id = f.income_id AND
      d9.marital_status_id = f.marital_status_id
      GROUP BY d1.calendar_year_name, d1.calendar_quarter_name, d2.all_products_name,
      d2.department_name, d2.category_name, d2.type_name, d3.all_customers_name,
      d3.region_name, d3.country_name, d3.state_province_name, d4.all_channels_name,
      d4.class_name, d4.channel_name, d5.all_ages_name, d5.age_name, d6.all_sizes_name,
      d6.household_size_name, d7.all_years_name, d7.years
      

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of October 25, 2015.

    Neural Network Models Using Oracle R Enterprise: SPARC T7-4 Beats 4-Chip x86 E7 v3

    Oracle's SPARC T7-4 server executing neural network algorithms using Oracle R Enterprise (ORE) is up to two times faster than a four-chip x86 E7 v3 server.

    • For a neural network with two hidden layers, 10-neuron with 5-neuron hyperbolic tangent, the SPARC T7-4 server is 1.5 times faster than a four-chip x86 T7 v3 server on calculation time.

    • For a neural network with two hidden layers, 20-neuron with 10-neuron hyperbolic tangent, the SPARC T7-4 server is 2.0 times faster than than a four-chip x86 T7 v3 server on calculation time.

    Performance Landscape

    Oracle Enterprise R Statistics in Oracle Database
    (250 million rows)
    Neural Network
    with Two Hidden Layers
    Elapsed Calculation Time SPARC Advantage
    4-chip x86 E7 v3 SPARC T7-4
    10-neuron + 5-neuron
    hyperbolic tangent
    520.1 (sec) 337.3 (sec) 1.5x
    20-neuron + 10-neuron
    hyperbolic tangent
    1128.4 (sec) 578.1 (sec) 2.0x

    Configuration Summary

    SPARC Configuration:

    SPARC T7-4
    4 x SPARC M7 processors (4.13 GHz)
    2 TB memory (64 x 32 GB dimms)
    Oracle Solaris 11.3
    Oracle Database 12c Enterprise Edition
    Oracle R Enterprise 1.5
    Oracle Solaris Studio 12.4 with 4/15 patch set

    x86 Configuration:

    Oracle Server X5-4
    4 x Intel Xeon Processor E7-8895 v3 (2.6 GHz)
    512 GB memory
    Oracle Linux 6.4
    Oracle Database 12c Enterprise Edition
    Oracle R Enterprise 1.5

    Storage Configuration:

    Oracle Server X5-2L
    2 x Intel Xeon Processor E5-2699 v3
    512 GB memory
    4 x 1.6 TB 2.5-inch NVMe PCIe 3.0 SSD
    2 x Sun Storage Dual 16Gb FC PCIe HBA
    Oracle Solaris 11.3

    Benchmark Description

    The benchmark is designed to run various statistical analyses using Oracle R Enterprise (ORE) with historical aviation data. The size of the benchmark data is about 35 GB, a single table holding 250 million rows. One of the most popular algorithms, neural network, has been used against the dataset to generate comparable results.

    The neural network algorithms support various features. In this workload, the following two neural network features have been used: neural net with two hidden layers 10-neuron with 5-neuron hyperbolic tangent and neural net with two hidden layers 20-neuron with 10-neuron hyperbolic tangent.

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 25 October 2015.

    SPECjEnterprise2010: SPARC T7-1 World Record with Single Application Server Using 1 to 4 Chips

    Oracle's SPARC T7-1 servers have set a world record for the SPECjEnterprise2010 benchmark for solutions using a single application server with one to four chips. The result of 25,818.85 SPECjEnterprise2010 EjOPS used two SPARC T7-1 servers, one server for the application tier and the other server for the database tier.

    • The SPARC T7-1 servers obtained a result of 25,093.06 SPECjEnterprise2010 EjOPS using encrypted data. This secured result used Oracle Advanced Security Transparent Data Encryption (TDE) for the application database tablespaces with the AES-256-CFB cipher. The network connection between the application server and the database server was also encrypted using the secure JDBC.

    • The SPARC T7-1 server solution delivered 34% more performance compared to the two-chip IBM x3650 M5 server result of 19,282.14 SPECjEnterprise2010 EjOPS.

    • The SPARC T7-1 server solution delivered 14% more performance compared to the four-chip IBM Power System S824 server result of 22,543.34 SPECjEnterprise2010 EjOPS.

    • The SPARC T7-1 server based results demonstrated 20% more performance compared to the Oracle Server X5-2 system result of 21,504.30 SPECjEnterprise2010 EjOPS. Oracle holds the top x86 two-chip application server SPECjEnterprise2010 result.

    • The application server used Oracle Fusion Middleware components including the Oracle WebLogic 12.1 application server and Java HotSpot(TM) 64-Bit Server VM on Solaris, version 1.8.0_60. The database server was configured with Oracle Database 12c Release 1.

    • For the secure result, the application data was encrypted in the Oracle database using the Oracle Advanced Security Transparent Data Encryption (TDE) feature. Hardware accelerated cryptography support in the SPARC M7 processor for the AES-256-CFB cipher was used to provide data security.

    • The benchmark performance using the secure SPARC T7-1 server configuration with encryption was less than 3% when compared to the peak result.

    • This result demonstrated less than 1 second average response times for all SPECjEnterprise2010 transactions and represents Jave EE 5.0 transactions generated by over 210,000 users.

    Performance Landscape

    Select single application server results. Complete benchmark results are at the SPEC website, SPECjEnterprise2010 Results.

    SPECjEnterprise2010 Performance Chart
    10/25/2015
    Submitter EjOPS* Java EE Server DB Server Notes
    Oracle 25,818.85 1 x SPARC T7-1
    1 x 4.13 GHz SPARC M7
    Oracle WebLogic 12c (12.1.3)
    1 x SPARC T7-1
    1 x 4.13 GHz SPARC M7
    Oracle Database 12c (12.1.0.2)
    -
    Oracle 25,093.06 1 x SPARC T7-1
    1 x 4.13 GHz SPARC M7
    Oracle WebLogic 12c (12.1.3)
    Network Data Encryption for JDBC
    1 x SPARC T7-1
    1 x 4.13 GHz SPARC M7
    Oracle Database 12c (12.1.0.2)
    Transparent Data Encryption
    Secure
    IBM 22,543.34 1 x IBM Power S824
    4 x 3.5 GHz POWER 8
    WebSphere Application Server V8.5
    1 x IBM Power S824
    4 x 3.5 GHz POWER 8
    IBM DB2 10.5 FP3
    -
    Oracle 21,504.30 1 x Oracle Server X5-2
    2 x 2.3 GHz Intel Xeon E5-2699 v3
    Oracle WebLogic 12c (12.1.3)
    1 x Oracle Server X5-2
    2 x 2.3 GHz Intel Xeon E5-2699 v3
    Oracle Database 12c (12.1.0.2)
    COD
    IBM 19,282.14 1 x System x3650 M5
    2 x 2.6 GHz Intel Xeon E5-2697 v3
    WebSphere Application Server V8.5
    1 x System x3850 X6
    4 x 2.8 GHz Intel Xeon E7-4890 v2
    IBM DB2 10.5 FP5
    -

    * SPECjEnterprise2010 EjOPS (bigger is better)

    The Cluster on Die (COD) mode is a BIOS setting that effectively splits the chip in half, making the operating system think it has twice as many chips as it does (in this case, four, 9 core chips). Intel has stated that COD is appropriate only for highly NUMA optimized workloads. Dell has shown that there is a 3.7x slower bandwidth to the other half of the chip split by COD.

    Configuration Summary

    Application Server:

    1 x SPARC T7-1 server, with
    1 x SPARC M7 processor (4.13 GHz)
    256 GB memory (16 x 16 GB)
    2 x 600 GB SAS HDD
    2 x 400 GB SAS SSD
    3 x Sun Dual Port 10 GbE PCIe 2.0 Networking card with Intel 82599 10 GbE Controller
    Oracle Solaris 11.3 (11.3.0.0.30)
    Oracle WebLogic Server 12c (12.1.3)
    Java HotSpot(TM) 64-Bit Server VM on Solaris, version 1.8.0_60

    Database Server:

    1 x SPARC T7-1 server, with
    1 x SPARC M7 processor (4.13 GHz)
    512 GB memory (16 x 32 GB)
    2 x 600 GB SAS HDD
    1 x Sun Dual Port 10 GbE PCIe 2.0 Networking card with Intel 82599 10 GbE Controller
    1 x Sun Storage 16 Gb Fibre Channel Universal HBA
    Oracle Solaris 11.3 (11.3.0.0.30)
    Oracle Database 12c (12.1.0.2)

    Storage Servers:

    1 x Oracle Server X5-2L (8-Drive), with
    2 x Intel Xeon Processor E5-2699 v3 (2.3 GHz)
    32 GB memory
    1 x Sun Storage 16 Gb Fibre Channel Universal HBA
    4 x 1.6 TB NVMe SSD
    2 x 600 GB SAS HDD
    Oracle Solaris 11.3 (11.3.0.0.30)
    1 x Oracle Server X5-2L (24-Drive), with
    2 x Intel Xeon Processor E5-2699 v3 (2.3 GHz)
    32 GB memory
    1 x Sun Storage 16 Gb Fibre Channel Universal HBA
    14 x 600 GB SAS HDD
    Oracle Solaris 11.3 (11.3.0.0.30)

    1 x Brocade 6510 16 Gb FC switch

    Benchmark Description

    SPECjEnterprise2010 is the third generation of the SPEC organization's J2EE end-to-end industry standard benchmark application. The SPECjEnterprise2010 benchmark has been designed and developed to cover the Java EE 5 specification's significantly expanded and simplified programming model, highlighting the major features used by developers in the industry today. This provides a real world workload driving the Application Server's implementation of the Java EE specification to its maximum potential and allowing maximum stressing of the underlying hardware and software systems,

    • The web zone, servlets, and web services
    • The EJB zone
    • JPA 1.0 Persistence Model
    • JMS and Message Driven Beans
    • Transaction management
    • Database connectivity
    Moreover, SPECjEnterprise2010 also heavily exercises all parts of the underlying infrastructure that make up the application environment, including hardware, JVM software, database software, JDBC drivers, and the system network.

    The primary metric of the SPECjEnterprise2010 benchmark is jEnterprise Operations Per Second (SPECjEnterprise2010 EjOPS). The primary metric for the SPECjEnterprise2010 benchmark is calculated by adding the metrics of the Dealership Management Application in the Dealer Domain and the Manufacturing Application in the Manufacturing Domain. There is NO price/performance metric in this benchmark.

    Key Points and Best Practices

    • Four Oracle WebLogic server instances on the SPARC T7-1 server were hosted in 4 separate Oracle Solaris Zones.
    • The Oracle WebLogic application servers were executed in the FX scheduling class to improve performance by reducing the frequency of context switches.
    • The Oracle log writer process was run in the RT scheduling class.

    See Also

    Disclosure Statement

    SPEC and the benchmark name SPECjEnterprise are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 10/25/2015. SPARC T7-1, 25,818.85 SPECjEnterprise2010 EjOPS (unsecure); SPARC T7-1, 25,093.06 SPECjEnterprise2010 EjOPS (secure); Oracle Server X5-2, 21,504.30 SPECjEnterprise2010 EjOPS (unsecure); IBM Power S824, 22,543.34 SPECjEnterprise2010 EjOPS (unsecure); IBM x3650 M5, 19,282.14 SPECjEnterprise2010 EjOPS (unsecure);

    SPECvirt_sc2013: SPARC T7-2 World Record for 2 and 4 Chip Systems

    Oracle has had a new result accepted by SPEC as of November 19, 2015. This new result may be found here.

    Oracle's SPARC T7-2 server delivered a world record SPECvirt_sc2013 result for systems with two to four chips.

    • The SPARC T7-2 server produced a result of 3026 @ 168 VMs SPECvirt_sc2013.

    • The two-chip SPARC T7-2 server beat the best two-chip x86 Intel E5-2699 v3 server results by nearly 1.9 times (Huawei FusionServer RH2288H V3, HP ProLiant DL360 Gen9).

    • The two-chip SPARC T7-2 server delivered nearly 2.2 times the performance of the four-chip IBM Power System S824 server solution which used 3.5 GHz POWER8 six core chips.

    • The SPARC T7-2 server running Oracle Solaris 11.3 operating system, utilizes embedded virtualization products as the Oracle Solaris 11 zones, which in turn provide a low overhead, flexible, scalable and manageable virtualization environment.

    • The SPARC T7-2 server result used Oracle VM Server for SPARC 3.3 and Oracle Solaris Zones providing a flexible, scalable and manageable virtualization environment.

    Performance Landscape

    Complete benchmark results are at the SPEC website, SPECvirt_sc2013 Results. The following table highlights the leading two-, and four-chip results for the benchmark, bigger is better.

    SPECvirt_sc2013
    Leading Two to Four-Chip Results
    System
    Processor
    Chips Result @ VMs Virtualization Software
    SPARC T7-2
    SPARC M7 (4.13 GHz, 32core)
    2 3026 @ 168 Oracle VM Server for SPARC 3.3
    Oracle Solaris Zones
    HP DL580 Gen9
    Intel E7-8890 v3 (2.5 GHz, 18core)
    4 3020 @ 168 Red Hat Enterprise Linux 7.1 KVM
    Lenovo System x3850 X6
    Intel E7-8890 v3 (2.5 GHz, 18core)
    4 2655 @ 147 Red Hat Enterprise Linux 6.6 KVM
    Huawei FusionServer RH2288H V3
    Intel E5-2699 v3 (2.3 GHz, 18core)
    2 1616 @ 95 Huawei FusionSphere V1R5C10
    HP DL360 Gen9
    Intel E5-2699 v3 (2.3 GHz, 18core)
    2 1614 @ 95 Red Hat Enterprise Linux 7.1 KVM
    IBM Power S824
    POWER8 (3.5 GHz, 6core)
    4 1370 @ 79 PowerVM Enterprise Edition 2.2.3

    Configuration Summary

    System Under Test Highlights:

    Hardware:
    1 x SPARC T7-2 server, with
    2 x 4.13 GHz SPARC M7
    1 TB memory
    2 Sun Dual Port 10GBase-T Adapter
    2 Sun Storage Dual 16 Gb Fibre Channel PCIe Universal HBA

    Software:
    Oracle Solaris 11.3
    Oracle VM Server for SPARC 3.3 (LDom)
    Oracle Solaris Zones
    Oracle iPlanet Web Server 7.0.20
    Oracle PHP 5.3.29
    Dovecot v2.2.18
    Oracle WebLogic Server Standard Edition Release 10.3.6
    Oracle Database 12c Enterprise Edition (12.1.0.2.0)
    Java HotSpot(TM) 64-Bit Server VM on Solaris, version 1.7.0_85-b15

    Storage:
    3 x Oracle Server X5-2L, with
    2 x Intel Xeon Processor E5-2630 v3 8-core 2.4 GHz
    32 GB memory
    4 x Oracle Flash Accelerator F160 PCIe Card
    Oracle Solaris 11.3

    1 x Oracle Server X5-2L, with
    2 x Intel Xeon Processor E5-2630 v3 8-core 2.4 GHz
    32 GB memory
    4 x Oracle Flash Accelerator F160 PCIe Card
    4x 400 GB SSDs
    Oracle Solaris 11.3

    Benchmark Description

    SPECvirt_sc2013 is SPEC's updated benchmark addressing performance evaluation of datacenter servers used in virtualized server consolidation. SPECvirt_sc2013 measures the end-to-end performance of all system components including the hardware, virtualization platform, and the virtualized guest operating system and application software. It utilizes several SPEC workloads representing applications that are common targets of virtualization and server consolidation. The workloads were made to match a typical server consolidation scenario of CPU resource requirements, memory, disk I/O, and network utilization for each workload. These workloads are modified versions of SPECweb2005, SPECjAppServer2004, SPECmail2008, and SPEC CPU2006. The client-side SPECvirt_sc2013 harness controls the workloads. Scaling is achieved by running additional sets of virtual machines, called "tiles", until overall throughput reaches a peak.

    Key Points and Best Practices

    • The SPARC T7-2 server running the Oracle Solaris 11.3, utilizes embedded virtualization products as the Oracle VM Server for SPARC and Oracle Solaris Zones, which provide a low overhead, flexible, scalable and manageable virtualization environment.

    • In order to provide a high level of data integrity and availability, all the benchmark data sets are stored on mirrored (RAID1) storage

    • Using Oracle VM Server for SPARC to bind the SPARC M7 processor with its local memory optimized system memory use in this virtual environment.

    See Also

    Disclosure Statement

    SPEC and the benchmark name SPECvirt_sc are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 10/25/2015. SPARC T7-2, SPECvirt_sc2013 3026@168 VMs; HP DL580 Gen9, SPECvirt_sc2013 3020@168 VMs; Lenovo x3850 X6; SPECvirt_sc2013 2655@147 VMs; Huawei FusionServer RH2288H V3, SPECvirt_sc2013 1616@95 VMs; HP ProLiant DL360 Gen9, SPECvirt_sc2013 1614@95 VMs; IBM Power S824, SPECvirt_sc2013 1370@79 VMs.

    Oracle Internet Directory: SPARC T7-2 World Record

    Oracle's SPARC T7-2 server running Oracle Internet Directory (OID, Oracle's LDAP Directory Server) on Oracle Solaris 11 on a virtualized processor configuration achieved a record result on the Oracle Internet Directory benchmark.

    • The SPARC T7-2 server, virtualized to use a single processor, achieved world record performance running Oracle Internet Directory benchmark with 50M users.

    • The SPARC T7-2 server and Oracle Internet Directory using Oracle Database 12c running on Oracle Solaris 11 achieved record result of 1.18M LDAP searches/sec with an average latency of 0.85 msec with 1000 clients.

    • The SPARC T7 server demonstrated 25% better throughput and 23% better latency for LDAP search/sec over similarly configured SPARC T5 server benchmark environment.

    • Oracle Internet Directory achieved near linear scalability on the virtualized single processor domain on the SPARC T7-2 server with 79K LDAP searches/sec with 2 cores to 1.18M LDAP searches/sec with 32 cores.

    • Oracle Internet Directory and the virtualized single processor domain on the SPARC T7-2 server achieved up to 22,408 LDAP modify/sec with an average latency of 2.23 msec for 50 clients.

    Performance Landscape

    A virtualized single SPARC M7 processor in a SPARC T7-2 server was used for the test results presented below. The SPARC T7-2 server and SPARC T5-2 server results were run as part of this benchmark effort. The remaining results were part of a previous benchmark effort.

    Oracle Internet Directory Tests
    System chips/
    cores
    Search Modify Add
    ops/sec lat (msec) ops/sec lat (msec) ops/sec lat (msec)
    SPARC T7-2 1/32 1,177,947 0.85 22,400 2.2 1,436 11.1
    SPARC T5-2 2/32 944,624 1.05 16,700 2.9 1,000 15.95
    SPARC T4-4 4/32 682,000 1.46 12,000 4.0 835 19.0

    Scaling runs were also made on the virtualized single processor domain on the SPARC T7-2 server.

    Scaling of Search Tests – SPARC T7-2, One Processor
    Cores Clients ops/sec Latency (msec)
    32 1000 1,177,947 0.85
    24 1000 863,343 1.15
    16 500 615,563 0.81
    8 500 280,029 1.78
    4 100 156,114 0.64
    2 100 79,300 1.26

    Configuration Summary

    System Under Test:

    SPARC T7-2
    2 x SPARC M7 processors, 4.13 GHz
    512 GB memory
    6 x 600 GB internal disks
    1 x Sun Storage ZS3-2 (used for database and log files)
    Flash storage (used for redo logs)
    Oracle Solaris 11.3
    Oracle Internet Directory 11g Release 1 PS7 (11.1.1.7.0)
    Oracle Database 12c Enterprise Edition 12.1.0.2 (64-bit)

    Benchmark Description

    Oracle Internet Directory (OID) is Oracle's LDAPv3 Directory Server. The throughput for five key operations are measured — Search, Compare, Modify, Mix and Add.

    LDAP Search Operations Test

    This test scenario involved concurrent clients binding once to OID and then performing repeated LDAP Search operations. The salient characteristics of this test scenario is as follows:

    • SLAMD SearchRate job was used.
    • BaseDN of the search is root of the DIT, the scope is SUBTREE, the search filter is of the form UID=, DN and UID are the required attribute.
    • Each LDAP search operation matches a single entry.
    • The total number concurrent clients was 1000 and were distributed amongst two client nodes.
    • Each client binds to OID once and performs repeated LDAP Search operations, each search operation resulting in the lookup of a unique entry in such a way that no client looks up the same entry twice and no two clients lookup the same entry and all entries are searched randomly.
    • In one run of the test, random entries from the 50 Million entries are looked up in as many LDAP Search operations.
    • Test job was run for 60 minutes.

    LDAP Compare Operations Test

    This test scenario involved concurrent clients binding once to OID and then performing repeated LDAP Compare operations on userpassword attribute. The salient characteristics of this test scenario is as follows:

    • SLAMD CompareRate job was used.
    • Each LDAP compare operation matches user password of user.
    • The total number concurrent clients was 1000 and were distributed amongst two client nodes.
    • Each client binds to OID once and performs repeated LDAP compare operations.
    • In one run of the test, random entries from the 50 Million entries are compared in as many LDAP compare operations.
    • Test job was run for 60 minutes.

    LDAP Modify Operations Test

    This test scenario consisted of concurrent clients binding once to OID and then performing repeated LDAP Modify operations. The salient characteristics of this test scenario is as follows:

    • SLAMD LDAP modrate job was used.
    • A total of 50 concurrent LDAP clients were used.
    • Each client updates a unique entry each time and a total of 50 Million entries are updated.
    • Test job was run for 60 minutes.
    • Value length was set to 11.
    • Attribute that is being modified is not indexed.

    LDAP Mixed Load Test

    The test scenario involved both the LDAP search and LDAP modify clients enumerated above.

    • The ratio involved 60% LDAP search clients, 30% LDAP bind and 10% LDAP modify clients.
    • A total of 1000 concurrent LDAP clients were used and were distributed on 2 client nodes.
    • Test job was run for 60 minutes.

    LDAP Add Load Test

    The test scenario involved concurrent clients adding new entries as follows.

    • Slamd standard add rate job is used.
    • A total of 500,000 entries were added.
    • A total of 16 concurrent LDAP clients were used.
    • Slamd add's inetorgperson objectclass entry with 21 attributes (includes operational attributes).

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 25 October 2015.

    Oracle Stream Explorer DDOS Attack: SPARC T7-4 World Record

    A single processor of Oracle's SPARC T7-4 server achieved a world record result running an Oracle Stream Explorer platform benchmark. The Oracle Stream Explorer platform is used to process multiple event streams to detect patterns and trends in real time. The benchmark detects malicious IP addresses that cause a distributed denial of service (DDOS) attack.

    • A single SPARC M7 processor of a SPARC T7-4 server running Oracle Stream Explorer achieved a throughput result of 1.505 million ops/sec.

    • The SPARC M7 processor achieved 2.9 times the throughput of an x86 Intel Xeon Processor E7-8895 v3 based server.

    Performance Landscape

    All of the following results were run as part of this benchmark effort.

    Oracle Stream Explorer Throughput Test
    One Processor Performance
    System Throughput
    SPARC T7-4 1.505 M ops/sec
    Oracle Server X5-4 0.522 M ops/sec

    Configuration Summary

    SPARC Server:

    SPARC T7-4
    4 x SPARC M7 processors
    1 TB memory
    Oracle Solaris 11.3
    Oracle Stream Explorer 11.1.1.7 (PS6)
    Oracle JDK 6

    x86 Server:

    Oracle Server X5-4
    4 x Intel Xeon Processor E7-8895 v3
    1 TB memory
    Oracle Solaris 11.3
    Oracle Stream Explorer 11.1.1.7 (PS6)
    Oracle JDK 6

    Benchmark Description

    The benchmark detects malicious IP addresses that cause a distributed denial of service (DDOS) attack on a system. The benchmark determines which IP address sent the most packets. The benchmark has a dedicated load generator program for each Oracle Stream Explorer platform instance.

    The Oracle Stream Explorer platform instance is always in a listening mode. When it receives data on its network socket, it starts incrementing the packet counter. Different Oracle Stream Explorer platform instances are deployed on different network sockets. The packet counter is printed out in regular intervals as the throughput for benchmarking purposes.

    Key Points and Best Practices

    • The load generator was run on the system under test. One processor was used for the event processing, the other processors were used for the load generation.

    • On the SPARC T7-4 server, three SPARC M7 processors were assigned the task of running the 200 load generators. This was accomplished using the psrset command.

    • On the Oracle Server X5-4 system, three Intel Xeon Processor E7-8895 v3 were assigned the task of running the 36 load generators.

    • Only 25 cores of the SPARC M7 processor were required to satisfy the workload. The 200 Oracle Stream Explorer applications were bound eight per core.

    • All 18 cores of the Intel Xeon Processor E7-8895 v3 were required to satisfy the workload. The 36 Oracle Stream Explorer applications were bound two per core.

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 25 October 2015.

    Oracle FLEXCUBE Universal Banking: SPARC T7-1 World Record

    Oracle's SPARC T7-1 servers running Oracle FLEXCUBE Universal Banking Release 12 along with Oracle Database 12c Enterprise Edition with Oracle Real Application Clusters on Oracle Solaris 11 produced record results for two processor solutions.

    • Two SPARC T7-1 servers each running Oracle FLEXCUBE Universal Banking Release 12 (v 12.0.1) and Oracle Real Application Clusters 12c database on Oracle Solaris 11 achieved record End of Year batch processing of 25 million accounts with 200 branches in 4 hrs 34 minutes (total of two processors).

    • A single SPARC T7-1 server running Oracle FLEXCUBE Universal Banking Release 12 processing 100 branches was able to complete the workload in similar time as the two node 200 branches End of Year workload, demonstrating good scaling of the application.

    • The customer representative workload for all 25 million accounts included saving accounts, current accounts, loans and TD accounts were created on the basis 25 million Customer IDs with 200 branches.

    • Oracle's SPARC M7 and T7 Servers running Oracle Solaris 11 with built-in Silicon Secured Memory with Oracle Database 12c can benefit global retail and corporate financial institutions who are running Oracle FLEXCUBE Universal Banking Release 12. The uniquely co-engineered Oracle software and hardware unlock unique agile capabilities demanded by modern business environments.

    • The SPARC T7-1 system and Oracle Solaris are able to provide a combination of uniquely essential characteristics that resonate with core values for a modern financial services institution.

    • The SPARC M7 processor based systems are capable of delivering higher performance and lower total cost of ownership (TCO) than older SPARC infrastructure, without introducing the unseen tax and risk of migrating applications away from older SPARC systems.

    Performance Landscape

    Oracle FLEXCUBE Universal Banking Release 12
    End of Year Batch Processing
    System Branches Time in Minutes
    2 x SPARC T7-1 200 274 (min)
    1 x SPARC T7-1 100 268 (min)

    Configuration Summary

    Systems Under Test:

    2 x SPARC T7-1 each with
    1 x SPARC M7 processor, 4.13 GHz
    256 GB memory
    Oracle Solaris 11.3 (11.3.0.27.0)
    Oracle Database 12c (RAC/ASM 12.1.0.2 BP7)
    Oracle FLEXCUBE Universal Banking Release 12

    Storage Configuration:

    Oracle ZFS Storage ZS4-4 appliance

    Benchmark Description

    The Oracle FLEXCUBE Universal Banking Release 12 benchmark models an actual customer bank with End of Cycle transaction batch jobs which typically execute during non-banking hours. This benchmark includes accrual for savings and term deposit accounts, interest capitalization for saving accounts, interest pay out for term deposit accounts and consumer load processing.

    This benchmark helps banks refine their infrastructure requirements for the volumes and scale of operations for business expansion. The end of cycle can be year, month or day, with year having the most processing followed by month and then day.

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 25 October 2015.

    PeopleSoft Human Capital Management 9.1 FP2: SPARC M7-8 World Record

    This result demonstrates how Oracle's SPARC M7-8 server using Oracle VM Server for SPARC (LDoms) provides mission critical enterprise virtualization.

    • The virtualized two-chip, 1 TB LDom of the SPARC M7-8 server set a world record two-chip PeopleSoft Human Capital Management (HCM) 9.1 FP2 benchmark result, supporting 35,000 HR Self-Service online users with response times under one second, while simultaneously running the Payroll batch workload.

    • The virtualized two-chip LDom of the SPARC M7-8 server demonstrated 4 times better Search and 6 times better Save average response times running nearly double the number of online users along with payroll batch, compared to the ten-chip x86 solution from Cisco.

    • Using only a single chip in the virtualized two-chip LDom on the SPARC M7-8 server, the batch-only run demonstrated 1.8 times better throughput (payments/hour) compared to a four-chip Cisco UCSB460 M4 server.

    • Using only a single chip in the virtualized two-chip LDom on the SPARC M7-8 server, the batch-only run demonstrated 2.3 times better throughput (payments/hour) compared to a nine-chip IBM zEnterprise z196 server (EC 2817-709, 9-way, 8943 MIPS).

    • This record result demonstrates that a two SPARC M7 processor LDom (in SPARC M7-8), can run the same number of online users as a dynamic domain (PDom) of eight SPARC M6 processors (in SPARC M6-32), with better online response times, batch elapsed times and batch throughput (payments/hour).

    • The SPARC M7-8 server provides enterprise applications high availability and security, where each application is executed on its own environment independent of the others.

    Performance Landscape

    The first table presents the combined results, running both the PeopleSoft HR Self-Service Online and Payroll Batch tests concurrently.

    PeopleSoft HR Self-Service Online And Payroll Batch Using Oracle Database 11g
    System
    Processors
    Chips
    Used
    Users Search/Save Batch Elapsed
    Time
    Batch Pay/Hr
    SPARC M7-8
    SPARC M7
    LDom1 2 35,000 0.67 sec/0.42 sec 22.71 min 1,322,272
    LDom2 2 35,000 0.85 sec/0.50 sec 22.96 min 1,307,875
    SPARC M6-32
    SPARC M6
    8 35,000 1.80 sec/1.12 sec 29.2 min 1,029,440
    Cisco 1 x B460 M4, 3 x B200 M3
    Intel E7-4890 v2, Intel E5-2697 v2
    10 18,000 2.70 sec/2.60 sec 21.70 min 1,383,816

    The following results are running only the Peoplesoft HR Self-Service Online test.

    PeopleSoft HR Self-Service Online Using Oracle Database 11g
    System
    Processors
    Chips
    Used
    Users Search/Save
    Avg Response Times
    SPARC M7-8
    SPARC M7
    LDom1 2 40,000 0.55 sec/0.33 sec
    LDom2 2 40,000 0.56 sec/0.32 sec
    SPARC M6-32
    SPARC M6
    8 40,000 2.73 sec/1.33 sec
    Cisco 1 x B460 M4, 3 x B200 M3
    Intel E7-4890 v2, Intel E5-2697 v2
    10 20,000 0.35 sec/0.17 sec

    The following results are running only the Peoplesoft Payroll Batch test. For the SPARC M7-8 server results, only one of the processors was used per LDom. This was accomplished using processor sets to further restrict the test to a single SPARC M7 processor.

    PeopleSoft Payroll Batch Using Oracle Database 11g
    System
    Processors
    Chips
    Used
    Batch Elapsed
    Time
    Batch Pay/Hr
    SPARC M7-8
    SPARC M7
    LDom1 1 13.06 min 2,299,296
    LDom2 1 12.85 min 2,336,872
    SPARC M6-32
    SPARC M6
    2 18.27 min 1,643,612
    Cisco UCS B460 M4
    Intel E7-4890 v2
    4 23.02 min 1,304,655
    IBM z196
    zEnterprise (5.2 GHz, 8943 MIPS)
    9 30.50 min 984,551

    Configuration Summary

    System Under Test:

    SPARC M7-8 server with
    8 x SPARC M7 processor (4.13 GHz)
    4 TB memory
    Virtualized as two Oracle VM Server for SPARC (LDom) each with
    2 x SPARC M7 processor (4.13 GHz)
    1 TB memory

    Storage Configuration:

    2 x Oracle ZFS Storage ZS3-2 appliance (DB Data) each with
    40 x 300 GB 10K RPM SAS-2 HDD,
    8 x Write Flash Accelerator SSD and
    2 x Read Flash Accelerator SSD 1.6TB SAS
    2 x Oracle Server X5-2L (DB redo logs & App object cache) each with
    2 x Intel Xeon Processor E5-2630 v3
    32 GB memory
    4 x 1.6 TB NVMe SSD

    Software Configuration:

    Oracle Solaris 11.3
    Oracle Database 11g Release 2 (11.2.0.3.0)
    PeopleSoft Human Capital Management 9.1 FP2
    PeopleSoft PeopleTools 8.52.03
    Oracle Java SE 6u32
    Oracle Tuxedo, Version 10.3.0.0, 64-bit, Patch Level 043
    Oracle WebLogic Server 11g (10.3.5)

    Benchmark Description

    The PeopleSoft Human Capital Management benchmark simulates thousands of online employees, managers and Human Resource administrators executing transactions typical of a Human Resources Self Service application for the Enterprise. Typical transactions are: viewing paychecks, promoting and hiring employees, updating employee profiles, etc. The database tier uses a database instance of about 500 GB in size, containing information for 500,480 employees. The application tier for this test includes web and application server instances, specifically Oracle WebLogic Server 11g, PeopleSoft Human Capital Management 9.1 FP2 and Oracle Java SE 6u32.

    Key Points and Best Practices

    In the HR online along with Payroll batch run, each LDom had one Oracle Solaris Zone of 7 cores containing the Web tier, two Oracle Solaris Zones of 16 cores each containing the Application tier and one Oracle Solaris Zone of 23 cores containing the Database tier. Two cores were dedicated to network and disk interrupt handling. In the HR online only run, each LDom had one Oracle Solaris Zone of 12 cores containing the Web tier, two Oracle Solaris Zones of 18 cores each containing the Application tier and one Oracle Solaris Zone of 14 cores containing the Database tier. 2 cores were dedicated to network and disk interrupt handling. In the Payroll batch only run, each LDom had one Oracle Solaris Zone of 31 cores containing the Database tier. 1 core was dedicated to disk interrupt handling.

    All database data files, recovery files and Oracle Clusterware files for the PeopleSoft test were created with the Oracle Automatic Storage Management (Oracle ASM) volume manager for the added benefit of the ease of management provided by Oracle ASM integrated storage management solution.

    In the application tier on each LDom, 5 PeopleSoft application domains with 350 application servers (70 per domain) were hosted in two separate Oracle Solaris Zones for a total of 10 domains with 700 application server processes.

    All PeopleSoft Application processes and the 32 Web Server JVM instances were executed in the Oracle Solaris FX scheduler class.

    See Also

    Disclosure Statement

    Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 10/25/2015.

    Oracle Communications ASAP – Telco Subscriber Activation: SPARC T7-2 World Record

    Oracle's SPARC T7-2 server delivered world record results on Oracle Communications ASAP. The SPARC T7-2 server ran Oracle Solaris 11 with Oracle Database 11g Release 2, Oracle WebLogic Server 12c and Oracle Communications ASAP version 7.2.

    • Running Oracle Communications ASAP, the SPARC T7-2 server delivered a world record result of 3,018 ASDLs/sec (atomic network activation actions).

    • Oracle's SPARC M7 processor delivered over 2.5 times the throughput per ASDL cost compared to the previous generation SPARC T5 processor.

    • The SPARC T7-2 server running a single instance of the Oracle Communications ASAP application, with both the application and database tiers consolidated onto a single machine, easily supported the service activation volumes of 3,018 ASDLs/sec which is representative of a typical mobile operator with more than 100 million subscribers.

    • Oracle Communications ASAP v7.2.0.4 delivered 35% higher throughput on the SPARC T7-2 server when compared to the SPARC T5-4 server.

    Performance Landscape

    All of the following results were run as part of this benchmark effort.

    ASAP 7.2.0.4 Test Results – 16 NEP
    Both tests used 1 cpu for App tier and 1 cpu for DB tier
    System ASDLs/sec CPU Usage CPU Cost per ASDL Cost Improvement Ratio
    SPARC T7-2 3,018.56 11.4% 1.10 2.6
    SPARC T5-4 2,238.97 29.6% 2.15 1.0

    CPU Cost per ASDL – computing cost per ASDL (smaller is better)
    Cost Improvement Ratio – improvement per cpu of SPARC T7-2 to SPARC T5-4

    Configuration Summary

    Hardware Configuration:

    SPARC T7-2 server
    2 x SPARC M7 processors (4.13 GHz)
    512 GB memory

    SPARC T5-4 server
    4 x SPARC T5 processors (3.6 GHz)
    512 GB memory

    Storage Configuration:

    Pillar Axiom

    Software Configuration:

    Oracle Communications ASAP 7.2.0.4.1
    Oracle Solaris 11.2
    Oracle Database 12c Release 12.1.0.1.0
    Oracle WebLogic Server 10.3.6.0
    Oracle JDK 7 update 75

    Benchmark Description

    Oracle Communications ASAP provides a convergent service activation platform that automatically activates customer services in a heterogeneous network and IT environment. It supports the activation of consumer and business services in fixed and mobile domains against network and IT applications.

    ASAP enables rapid service design and network technology introduction by means of its metadata-driven architecture, design-time configuration environment, and catalog of pre-built activation cartridges to reduce deployment time, cost, and risk. The application has been deployed for mobile (3G, 4G and M2M) services and fixed multi-play (broadband, voice, video, and IT) services in telecommunications, cable and satellite environments as well as for business voice, data, and IT cloud services.

    It may be deployed in a fully integrated manner as part of the Oracle Communications Service Fulfillment solution or directly integrated with third- party upstream systems. Market-proven for high-volume performance and scalability, Oracle Communications ASAP is deployed by more than 75 service providers worldwide and activates services for approximately 250 million subscribers globally.

    The throughput of ASAP is measured in atomic actions per second (or ASDLs/sec). An atomic action is a single command or operation that can be executed on a network element. Atomic actions are grouped together to form a common service action, where each service action typically relates to an orderable item, such as "GSM voice" or "voice mail" or "GSM data". One or more service actions are invoked by an order management system via an activation work order request.

    The workload resembles a typical mobile order to activate a GSM subscriber. A single service action to add a subscriber consists of seven atomic actions where each atomic action executes a command on a network element. Each network element was serviced by a dedicated Network Element Processor (NEP). The ASAP benchmark can vary the number of NEPs, which correlate to the complexity of a Telco operator's environment.

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 25 October 2015.

    Oracle E-Business Payroll Batch Extra-Large: SPARC T7-1 World Record

    Oracle's SPARC T7-1 server set a world record running the Oracle E-Business Suite 12.1.3 Standard Extra-Large (250,000 Employees) Payroll (Batch) workload.

    • The SPARC T7-1 server produced a world record result of 1,527,494 employee records processed per hour (9.82 min elapsed time) on the Oracle E-Business Suite R12 (12.1.3) Extra-Large Payroll (Batch) benchmark.

    • The SPARC T7-1 server equipped with one 4.13 GHz SPARC M7 processor, demonstrated 36% better hourly employee throughput compared to a two-chip Cisco UCS B200 M4 (Intel Xeon E5-2697 v3).

    • The SPARC T7-1 server equipped with one 4.13 GHz SPARC M7 processor, demonstrated 40% better hourly employee throughput compared to two-chip IBM S824 (POWER8 using 12 cores total).

    Performance Landscape

    This is the world record result for the Payroll Extra-Large model using Oracle E-Business 12.1.3 workload.

    Batch Workload: Payroll Extra-Large Model
    System Processor Employees/Hr Elapsed Time
    SPARC T7-1 1 x SPARC M7 (4.13 GHz) 1,527,494 9.82 minutes
    Cisco UCS B200 M4 2 x Intel Xeon Processor E5-2697 v3 1,125,281 13.33 minutes
    IBM S824 2 x POWER8 (3.52 GHz) 1,090,909 13.75 minutes
    Cisco UCS B200 M3 2 x Intel Xeon Processor E5-2697 v2 1,017,639 14.74 minutes
    Cisco UCS B200 M3 2 x Intel Xeon Processor E5-2690 839,865 17.86 minutes
    Sun Server X3-2L 2 x Intel Xeon Processor E5-2690 789,473 19.00 minutes

    Configuration Summary

    Hardware Configuration:

    SPARC T7-1 server
    1 x SPARC M7 processor (4.13 GHz)
    256 GB memory (16 x 16 GB)
    Oracle ZFS Storage ZS3-2 appliance (DB Data storage) with
    40 x 900 GB 10K RPM SAS-2 HDD,
    8 x Write Flash Accelerator SSD and
    2 x Read Flash Accelerator SSD 1.6 TB SAS
    Oracle Flash Accelerator F160 PCIe Card (1.6 TB NVMe for DB Log storage)

    Software Configuration:

    Oracle Solaris 11.3
    Oracle E-Business Suite R12 (12.1.3)
    Oracle Database 11g (11.2.0.3.0)

    Benchmark Description

    The Oracle E-Business Suite Standard R12 Benchmark combines online transaction execution by simulated users with concurrent batch processing to model a typical scenario for a global enterprise. This benchmark ran one Batch component, Payroll, in the Extra-Large size.

    Results can be published in four sizes and use one or more online/batch modules

    • X-large: Maximum online users running all business flows between 10,000 to 20,000; 750,000 order to cash lines per hour and 250,000 payroll checks per hour.
      • Order to Cash Online — 2400 users
        • The percentage across the 5 transactions in Order Management module is:
          • Insert Manual Invoice — 16.66%
          • Insert Order — 32.33%
          • Order Pick Release — 16.66%
          • Ship Confirm — 16.66%
          • Order Summary Report — 16.66%
      • HR Self-Service — 4000 users
      • Customer Support Flow — 8000 users
      • Procure to Pay — 2000 users
    • Large: 10,000 online users; 100,000 order to cash lines per hour and 100,000 payroll checks per hour.
    • Medium: up to 3000 online users; 50,000 order to cash lines per hour and 10,000 payroll checks per hour.
    • Small: up to 1000 online users; 10,000 order to cash lines per hour and 5,000 payroll checks per hour.

    Key Points and Best Practices

    • All system optimizations are in the published report which is referenced in the See Also section below.

    See Also

    Disclosure Statement

    Oracle E-Business X-Large Payroll Batch workload, SPARC T7-1, 4.13 GHz, 1 chip, 32 cores, 256 threads, 256 GB memory, elapsed time 9.82 minutes, 1,527,494 hourly employee throughput, Oracle Solaris 11.3, Oracle E-Business Suite 12.1.3, Oracle Database 11g Release 2, Results as of 10/25/2015.

    Oracle E-Business Suite Applications R12.1.3 (OLTP X-Large): SPARC M7-8 World Record

    Oracle's SPARC M7-8 server, using a four-chip Oracle VM Server for SPARC (LDom) virtualized server, produced a world record 20,000 users running the Oracle E-Business OLTP X-Large benchmark. The benchmark runs five Oracle E-Business online workloads concurrently: Customer Service, iProcurement, Order Management, Human Resources Self-Service, and Financials.

    • The virtualized four-chip LDom on the SPARC M7-8 was able to handle more users than the previous best result which used eight processors of Oracle's SPARC M6-32 server.

    • The SPARC M7-8 server using Oracle VM Server for SPARC provides enterprise applications high availability, where each application is executed on its own environment, insulated and independent of the others.

    Performance Landscape

    Oracle E-Business (3-tier) OLTP X-Large Benchmark
    System Chips Total Online Users Weighted Average
    Response Time (sec)
    90th Percentile
    Response Time (s)
    SPARC M7-8 4 20,000 0.70 1.13
    SPARC M6-32 8 18,500 0.61 1.16

    Break down of the total number of users by component.

    Users per Component
    Component SPARC M7-8 SPARC M6-32
    Total Online Users 20,000 users 18,500 users
    HR Self-Service
    Order-to-Cash
    iProcurement
    Customer Service
    Financial
    5000 users
    2500 users
    2700 users
    7000 users
    2800 users
    4000 users
    2300 users
    2400 users
    7000 users
    2800 users

    Configuration Summary

    System Under Test:

    SPARC M7-8 server
    8 x SPARC M7 processors (4.13 GHz)
    4 TB memory
    2 x 600 GB SAS-2 HDD
    using a Logical Domain with
    4 x SPARC M7 processors (4.13 GHz)
    2 TB memory
    2 x Sun Storage Dual 16Gb Fibre Channel PCIe Universal HBA
    2 x Sun Dual Port 10GBase-T Adapter
    Oracle Solaris 11.3
    Oracle E-Business Suite 12.1.3
    Oracle Database 11g Release 2

    Storage Configuration:

    4 x Oracle ZFS Storage ZS3-2 appliances each with
    2 x Read Flash Accelerator SSD
    1 x Storage Drive Enclosure DE2-24P containing:
    20 x 900 GB 10K RPM SAS-2 HDD
    4 x Write Flash Accelerator SSD
    1 x Sun Storage Dual 8Gb FC PCIe HBA
    Used for Database files, Zones OS, EBS Mid-Tier Apps software stack
    and db-tier Oracle Server
    2 x Sun Server X4-2L server with
    2 x Intel Xeon Processor E5-2650 v2
    128 GB memory
    1 x Sun Storage 6Gb SAS PCIe RAID HBA
    4 x 400 GB SSD
    14 x 600 GB HDD
    Used for Redo log files, db backup storage.

    Benchmark Description

    The Oracle E-Business OLTP X-Large benchmark simulates thousands of online users executing transactions typical of an internal Enterprise Resource Processing, simultaneously executing five application modules: Customer Service, Human Resources Self Service, iProcurement, Order Management and Financial.

    Each database tier uses a database instance of about 600 GB in size, supporting thousands of application users, accessing hundreds of objects (tables, indexes, SQL stored procedures, etc.).

    Key Points and Best Practices

    This test demonstrates virtualization technologies running concurrently various Oracle multi-tier business critical applications and databases on four SPARC M7 processors contained in a single SPARC M7-8 server supporting thousands of users executing a high volume of complex transactions with constrained (<1 sec) weighted average response time.

    The Oracle E-Business LDom is further configured using Oracle Solaris Zones.

    This result of 20,000 users was achieved by load balancing the Oracle E-Business Suite Applications 12.1.3 five online workloads across two Oracle Solaris processor sets and redirecting all network interrupts to a dedicated third processor set.

    Each applications processor set (set-1 and set-2) was running concurrently two Oracle E-Business Suite Application servers and two database servers instances, each within its own Oracle Solaris Zone (4 x Zones per set).

    Each application server network interface (to a client) was configured to map with the locality group associated to the CPUs processing the related workload, to guarantee memory locality of network structures and application servers hardware resources.

    All external storage was connected with at least two paths to the host multipath-capable fibre channel controller ports and Oracle Solaris I/O multipathing feature was enabled.

    See Also

    Disclosure Statement

    Oracle E-Business Suite R12 extra-large multiple-online module benchmark, SPARC M7-8, SPARC M7, 4.13 GHz, 4 chips, 128 cores, 1024 threads, 2 TB memory, 20,000 online users, average response time 0.70 sec, 90th percentile response time 1.13 sec, Oracle Solaris 11.3, Oracle Solaris Zones, Oracle VM Server for SPARC, Oracle E-Business Suite 12.1.3, Oracle Database 11g Release 2, Results as of 10/25/2015.

    Oracle E-Business Order-To-Cash Batch Large: SPARC T7-1 World Record

    Oracle's SPARC T7-1 server set a world record running the Oracle E-Business Suite 12.1.3 Standard Large (100,000 Order/Inventory Lines) Order-To-Cash (Batch) workload.

    • The SPARC T7-1 server produced a world record hourly order line throughput of 273,973 per hour (21.90 min elapsed time) on the Oracle E-Business Suite R12 (12.1.3) Large Order-To-Cash (Batch) benchmark using a SPARC T7-1 server for the database and application tiers running Oracle Database 11g on Oracle Solaris 11.

    • The SPARC T7-1 server demonstrated 12% better hourly order line throughput compared to a two-chip Cisco UCS B200 M4 (Intel Xeon Processor E5-2697 v3).

    Performance Landscape

    Results for the Oracle E-Business 12.1.3 Order-To-Cash Batch Large model workload.

    Batch Workload: Order-To-Cash Large Model
    System CPU Employees/Hr Elapsed Time (min)
    SPARC T7-1 1 x SPARC M7 processor 273,973 21.90
    Cisco UCS B200 M4 2 x Intel Xeon Processor E5-2697 v3 243,803 24.61
    Cisco UCS B200 M3 2 x Intel Xeon Processor E5-2690 232,739 25.78

    Configuration Summary

    Hardware Configuration:

    SPARC T7-1 server with
    1 x SPARC M7 processor (4.13 GHz)
    256 GB memory (16 x 16 GB)
    Oracle ZFS Storage ZS3-2 appliance (DB Data storage) with
    40 x 900 GB 10K RPM SAS-2 HDD,
    8 x Write Flash Accelerator SSD and
    2 x Read Flash Accelerator SSD 1.6TB SAS
    Oracle Flash Accelerator F160 PCIe Card (1.6 TB NVMe for DB Log storage)

    Software Configuration:

    Oracle Solaris 11.3
    Oracle E-Business Suite R12 (12.1.3)
    Oracle Database 11g (11.2.0.3.0)

    Benchmark Description

    The Oracle E-Business Suite Standard R12 Benchmark combines online transaction execution by simulated users with concurrent batch processing to model a typical scenario for a global enterprise. This benchmark ran one Batch component, Order-To-Cash, in the Large size.

    Results can be published in four sizes and use one or more online/batch modules

    • X-large: Maximum online users running all business flows between 10,000 to 20,000; 750,000 order to cash lines per hour and 250,000 payroll checks per hour.
      • Order to Cash Online — 2400 users
        • The percentage across the 5 transactions in Order Management module is:
          • Insert Manual Invoice — 16.66%
          • Insert Order — 32.33%
          • Order Pick Release — 16.66%
          • Ship Confirm — 16.66%
          • Order Summary Report — 16.66%
      • HR Self-Service — 4000 users
      • Customer Support Flow — 8000 users
      • Procure to Pay — 2000 users
    • Large: 10,000 online users; 100,000 order to cash lines per hour and 100,000 payroll checks per hour.
    • Medium: up to 3000 online users; 50,000 order to cash lines per hour and 10,000 payroll checks per hour.
    • Small: up to 1000 online users; 10,000 order to cash lines per hour and 5,000 payroll checks per hour.

    Key Points and Best Practices

    • All system optimizations are in the published report, find link in See Also section below.

    See Also

    Disclosure Statement

    Oracle E-Business Large Order-To-Cash Batch workload, SPARC T7-1, 4.13 GHz, 1 chip, 32 cores, 256 threads, 256 GB memory, elapsed time 21.90 minutes, 273,973 hourly order line throughput, Oracle Solaris 11.3, Oracle E-Business Suite 12.1.3, Oracle Database 11g Release 2, Results as of 10/25/2015.

    PeopleSoft Enterprise Financials 9.2: SPARC T7-2 World Record

    Oracle's SPARC T7-2 server achieved world record performance being the first to publish on Oracle's PeopleSoft Enterprise Financials 9.2 benchmark. This result was obtained using one Oracle VM Server for SPARC (LDom) virtualized system configured with a single SPARC M7 processor.

    • The single processor LDom on the SPARC T7-2 server achieved world record performance executing 200 million Journal Lines in 18.60 minutes.

    • The single processor LDom on the SPARC T7-2 server was able to process General Ledger Journal Edit and Post batch jobs at 10,752,688 journal lines/min which reflects a large customer environment that utilizes a back-end database of nearly 1.0 TB performing highly competitive journal processing for Ledger.

    Performance Landscape

    Results are presented for PeopleSoft Financials Benchmark 9.2. Results obtained with PeopleSoft Financials Benchmark 9.2 are not comparable to the the previous version of the benchmark, PeopleSoft Financials Benchmark 9.1, due to significant change in data model and supports only batch.

    PeopleSoft Financials Benchmark, Version 9.2
    Solution Under Test Batch Journal lines/min
    SPARC T7-2 (using 1 x SPARC M7, 4.13 GHz) 18.60 min 10,752,688

    Configuration Summary

    System:

    SPARC T7-2 server with
    2 x SPARC M7 processors
    1 TB memory
    4 x Oracle Flash Accelerator F160 PCIe Card (DB Redo, DB undo and DB Data)
    4 x 600 GB internal disks
    Oracle Solaris 11.3
    Oracle Database 11g (11.2.0.4)
    PeopleSoft Financials (9.20.348)
    PeopleSoft PeopleTools (8.53.09)
    Java HotSpot 64-Bit Server VM (build 1.7.0_45-b18)
    Oracle Tuxedo, Version 11.1.1.3.0, 64-bit
    Oracle WebLogic Server 11g (10.3.6)

    LDom Under Test:

    Oracle VM Server for SPARC (LDom) virtualized server (APP & DB Tier)
    1 x SPARC M7 processor
    512 GB memory

    Benchmark Description

    The PeopleSoft Enterprise Financials 9.2 benchmark emulates a large enterprise that processes and validates a large number of financial journal transactions before posting the journal entry to the ledger. The validation process certifies that the journal entries are accurate, ensuring that ChartFields values are valid, debits and credits equal out, and inter/intra-units are balanced. Once validated, the entries are processed, ensuring that each journal line posts to the correct target ledger, and then changes the journal status to posted. In this benchmark, the Journal Edit & Post is set up to edit and post both Inter-Unit and Regular multi-currency journals. The benchmark processes 200 million journal lines using AppEngine for edits and Cobol for post processes.

    Key Points and Best Practices

    • The PeopleSoft Enterprise Financials 9.2 Batch benchmark ran on a one chip LDom consisting of 32 cores, each core had 8 threads. All total there were 256 virtual processors.

    • The LDom contained two Oracle Solaris Zones: a database tier zone and an application tier zone. The application tier zone consisted of 1 core with 8 virtual processors. The database tier zone consisted of 244 virtual processors from 31 cores. The remaining four virtual processors were dedicated to network and disk interrupt handling.

    • Inside of the database tier zone, the database log writer ran under 4 virtual processors and eight virtual processors were dedicated to four database writers.

    • There were 160 PeopleSoft Application instance processes running 320 streams of PeopleSoft Financial workload in the Oracle Solaris Fixed Priority FX class.

    See Also

    Disclosure Statement

    Copyright 2015, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 25 October 2015.

    SAP Two-Tier Standard Sales and Distribution SD Benchmark: SPARC T7-2 World Record 2 Processors

    Oracle's SPARC T7-2 server produces a world record result for 2-processors on the SAP two-tier Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement Package 5 for SAP ERP 6.0 (2 chips / 64 cores / 512 threads).

    • The SPARC T7-2 server achieved 30,800 SAP SD benchmark users running the two-tier SAP Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement Package 5 for SAP ERP 6.0.

    • The SPARC T7-2 server achieved 1.9 times more users than the Dell PowerEdge R730 server result.

    • The SPARC T7-2 server achieved 1.5 times more users than the IBM Power System S824 server result.

    • The SPARC T7-2 server achieved 1.9 times more users than the HP ProLiant DL380 Gen9 server result.

    • The SPARC T7-2 server result was run with Oracle Solaris 11 and used Oracle Database 12c.

    Performance Landscape

    SAP-SD 2-tier performance table in decreasing performance order for leading two-processor systems and four-processor IBM Power System S824 server, with SAP ERP 6.0 Enhancement Package 5 for SAP ERP 6.0 results (current version of the benchmark as of May, 2012).

    SAP SD Two-Tier Benchmark
    System
    Processor
    OS
    Database
    Users Resp Time
    (sec)
    Version Cert#
    SPARC T7-2
    2 x SPARC M7 (2x 32core)
    Oracle Solaris 11
    Oracle Database 12c
    30,800 0.96 EHP5 2015050
    IBM Power S824
    4 x POWER8 (4x 6core)
    AIX 7
    DB2 10.5
    21,212 0.98 EHP5 2014016
    Dell PowerEdge R730
    2 x Intel E5-2699 v3 (2x 18core)
    Red Hat Enterprise Linux 7
    SAP ASE 16
    16,500 0.99 EHP5 2014033
    HP ProLiant DL380 Gen9
    2 x Intel E5-2699 v3 (2x 18core)
    Red Hat Enterprise Linux 6.5
    SAP ASE 16
    16,101 0.99 EHP5 2014032

    Version – Version of SAP, EHP5 refers to SAP ERP 6.0 Enhancement Package 5 for SAP ERP 6.0

    Number of cores presented are per chip, to get system totals, multiple by the number of chips.

    Complete benchmark results may be found at the SAP benchmark website http://www.sap.com/benchmark.

    Configuration Summary and Results

    Database/Application Server:

    1 x SPARC T7-2 server with
    2 x SPARC M7 processors (4.13 GHz, total of 2 processors / 64 cores / 512 threads)
    1 TB memory
    Oracle Solaris 11.3
    Oracle Database 12c

    Database Storage:
    3 x Sun Server X3-2L each with
    2 x Intel Xeon Processors E5-2609 (2.4 GHz)
    16 GB memory
    4 x Sun Flash Accelerator F40 PCIe Card
    12 x 3 TB SAS disks
    Oracle Solaris 11

    REDO log Storage:
    1 x Pillar FS-1 Flash Storage System, with
    2 x FS1-2 Controller (Netra X3-2)
    2 x FS1-2 Pilot (X4-2)
    4 x DE2-24P Disk enclosure
    96 x 300 GB 10000 RPM SAS Disk Drive Assembly

    Certified Results (published by SAP)

    Number of SAP SD benchmark users: 30,800
    Average dialog response time: 0.96 seconds
    Throughput:
      Fully processed order line items per hour: 3,372,000
      Dialog steps per hour: 10,116,000
      SAPS: 168,600
    Average database request time (dialog/update): 0.022 sec / 0.047 sec
    SAP Certification: 2015050

    Benchmark Description

    The SAP Standard Application SD (Sales and Distribution) Benchmark is an ERP business test that is indicative of full business workloads of complete order processing and invoice processing, and demonstrates the ability to run both the application and database software on a single system. The SAP Standard Application SD Benchmark represents the critical tasks performed in real-world ERP business environments.

    SAP is one of the premier world-wide ERP application providers, and maintains a suite of benchmark tests to demonstrate the performance of competitive systems on the various SAP products.

    See Also

    Disclosure Statement

    Two-tier SAP Sales and Distribution (SD) standard application benchmarks, SAP Enhancement Package 5 for SAP ERP 6.0 as of 10/23/15:

    SPARC T7-2 (2 processors, 64 cores, 512 threads) 30,800 SAP SD users, 2 x 4.13 GHz SPARC M7, 1 TB memory, Oracle Database 12c, Oracle Solaris 11, Cert# 2015050.
    IBM Power System S824 (4 processors, 24 cores, 192 threads) 21,212 SAP SD users, 4 x 3.52 GHz POWER8, 512 GB memory, DB2 10.5, AIX 7, Cert#2014016
    Dell PowerEdge R730 (2 processors, 36 cores, 72 threads) 16,500 SAP SD users, 2 x 2.3 GHz Intel Xeon Processor E5-2699 v3 256 GB memory, SAP ASE 16, RHEL 7, Cert#2014033
    HP ProLiant DL380 Gen9 (2 processors, 36 cores, 72 threads) 16,101 SAP SD users, 2 x 2.3 GHz Intel Xeon Processor E5-2699 v3 256 GB memory, SAP ASE 16, RHEL 6.5, Cert#2014032

    SAP, R/3, reg TM of SAP AG in Germany and other countries. More info www.sap.com/benchmark

    Friday Apr 03, 2015

    Oracle Server X5-2 Produces World Record 2-Chip Single Application Server SPECjEnterprise2010 Result

    Two Oracle Server X5-2 systems, using the Intel Xeon E5-2699 v3 processor, produced a World Record x86 two-chip single application server SPECjEnterprise2010 benchmark result of 21,504.30 SPECjEnterprise2010 EjOPS. One Oracle Server X5-2 ran the application tier and the second Oracle Server X5-2 was used for the database tier.

    • The Oracle Server X5-2 system demonstrated 11% better performance when compared to the IBM X3650 M5 server result of 19,282.14 SPECjEnterprise2010 EjOPS.

    • The Oracle Server X5-2 system demonstrated 1.9x better performance when compared to the previous generation Sun Server X4-2 server result of 11,259.88 SPECjEnterprise2010 EjOPS.

    • This result used Oracle WebLogic Server 12c, Java HotSpot(TM) 64-Bit Server 1.8.0_40 Oracle Database 12c, and Oracle Linux.

    Performance Landscape

    Complete benchmark results are at the SPEC website, SPECjEnterprise2010 Results. The table below shows the top single application server, two-chip x86 results.

    SPECjEnterprise2010 Performance Chart
    as of 4/1/2015
    Submitter EjOPS* Application Server Database Server
    Oracle 21,504.30 1x Oracle Server X5-2
    2x 2.3 GHz Intel Xeon E5-2699 v3
    Oracle WebLogic 12c (12.1.3)
    1x Oracle Server X5-2
    2x 2.3 GHz Intel Xeon E5-2699 v3
    Oracle Database 12c (12.1.0.2)
    IBM 19,282.14 1x IBM X3650 M5
    2x 2.6 GHz Intel Xeon E5-2697 v3
    WebSphere Application Server V8.5
    1x IBM X3850 X6
    4x 2.8 GHz Intel Xeon E7-4890 v2
    IBM DB2 10.5
    Oracle 11,259.88 1x Sun Server X4-2
    2x 2.7 GHz Intel Xeon E5-2697 v2
    Oracle WebLogic 12c (12.1.2)
    1x Sun Server X4-2L
    2x 2.7 GHz Intel Xeon E5-2697 v2
    Oracle Database 12c (12.1.0.1)

    * SPECjEnterprise2010 EjOPS, bigger is better.

    Configuration Summary

    Application Server:

    1 x Oracle Server X5-2
    2 x 2.3 GHz Intel Xeon E5-2699 v3 processors
    256 GB memory
    3 x 10 GbE NIC
    Oracle Linux 6 Update 5 (kernel-2.6.39-400.243.1.el6uek.x86_64)
    Oracle WebLogic Server 12c (12.1.3)
    Java HotSpot(TM) 64-Bit Server VM on Linux, version 1.8.0_40 (Java SE 8 Update 40)
    BIOS SW 1.2

    Database Server:

    1 x Oracle Server X5-2
    2 x 2.3 GHz Intel Xeon E5-2699 v3 processors
    512 GB memory
    2 x 10 GbE NIC
    1 x 16 Gb FC HBA
    2 x Oracle Server X5-2L Storage
    Oracle Linux 6 Update 5 (kernel-3.8.13-16.2.1.el6uek.x86_64)
    Oracle Database 12c Enterprise Edition Release 12.1.0.2

    Benchmark Description

    SPECjEnterprise2010 is the third generation of the SPEC organization's J2EE end-to-end industry standard benchmark application. The SPECjEnterprise2010 benchmark has been designed and developed to cover the Java EE 5 specification's significantly expanded and simplified programming model, highlighting the major features used by developers in the industry today. This provides a real world workload driving the Application Server's implementation of the Java EE specification to its maximum potential and allowing maximum stressing of the underlying hardware and software systems.

    The workload consists of an end to end web based order processing domain, an RMI and Web Services driven manufacturing domain and a supply chain model utilizing document based Web Services. The application is a collection of Java classes, Java Servlets, Java Server Pages, Enterprise Java Beans, Java Persistence Entities (pojo's) and Message Driven Beans.

    The SPECjEnterprise2010 benchmark heavily exercises all parts of the underlying infrastructure that make up the application environment, including hardware, JVM software, database software, JDBC drivers, and the system network.

    The primary metric of the SPECjEnterprise2010 benchmark is jEnterprise Operations Per Second ("SPECjEnterprise2010 EjOPS"). This metric is calculated by adding the metrics of the Dealership Management Application in the Dealer Domain and the Manufacturing Application in the Manufacturing Domain. There is no price/performance metric in this benchmark.

    Key Points and Best Practices

    • Four Oracle WebLogic server instances were started using numactl binding 2 instances per chip.
    • Four Oracle database listener processes were started, 2 processes bound per processor.
    • Additional tuning information is in the report at http://spec.org.
    • COD (Cluster on Die) is enabled in the BIOS on the application server.

    See Also

    Disclosure Statement

    SPEC and the benchmark name SPECjEnterprise are registered trademarks of the Standard Performance Evaluation Corporation. Oracle Server X5-2, 21,504.30 SPECjEnterprise2010 EjOPS; IBM System X3650 M5, 19,282.14 SPECjEnterprise2010 EjOPS. Sun Server X4-2, 11,259.88 SPECjEnterprise2010 EjOPS; Results from www.spec.org as of 4/1/2015.

    Thursday Mar 27, 2014

    SPARC M6-32 Produces SAP SD Two-Tier Benchmark World Record for 32-Processor Systems

    Oracle's SPARC M6-32 server produced a world record result for 32-processors on the SAP two-tier Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement Package 5 for SAP ERP 6.0 (32 chips / 384 cores / 3072 threads).

    • SPARC M6-32 server achieved 140,000 SAP SD benchmark users with a low average dialog response time of 0.58 seconds running the SAP two-tier Sales and Distribution (SD) Standard Application Benchmark using SAP Enhancement package 5 for SAP ERP 6.0.

    • The SPARC M6-32 delivered 2.5 times more users than the IBM Power 780 result using SAP Enhancement Package 5 for SAP ERP 6.0. The IBM result also had 1.7 times worse average dialog response time compared to the SPARC M6-32 server result.

    • The SPARC M6-32 delivered 3.0 times more users than the Fujitsu PRIMEQUEST 2800E (with Intel Xeon E7-8890 v2 processors) result. The Fujitsu result also had 1.7 times worse average dialog response time compared to the SPARC M6-32 server result.

    • The SPARC M6-32 server solution was run with Oracle Solaris 11 and used Oracle Database 11g.

    Performance Landscape

    SAP-SD 2-Tier Performance Table (in decreasing performance order). With SAP ERP 6.0 Enhancement Package 4 for SAP ERP 6.0 (Old version of the benchmark, obsolete at the end of April, 2012), and SAP ERP 6.0 Enhancement Package 5 for SAP ERP 6.0 results (current version of the benchmark as of May, 2012).

    System
    Processor
    Ch / Co / Th — Memory
    OS
    Database
    Users Resp Time
    (sec)
    Version Cert#
    Fujitsu SPARC M10-4S
    SPARC64 X @3.0 GHz
    40 / 640 / 1280 — 10 TB
    Solaris 11
    Oracle 11g
    153,000 0.87 EHP5 2013014
    SPARC M6-32 Server
    SPARC M6 @3.6 GHz
    32 / 384 / 3072 — 16 TB
    Solaris 11
    Oracle 11g
    140,000 0.58 EHP5 2014008
    IBM Power 795
    POWER7 @4 GHz
    32 / 256 / 1024 — 4 TB
    AIX 7.1
    DB2 9.7
    126,063 0.98 EHP4 2010046
    IBM Power 780
    POWER7+ @3.72 GHz
    12 / 96 / 834 — 1536 GB
    AIX 7.1
    DB2 10
    57,024 0.98 EHP5 2012033
    Fujitsu PRIMEQUEST 2800E
    Intel Xeon E7-8890 v2 @2.8 GHz
    8 / 120 / 240 — 1024 GB
    Windows Server 2012 SE
    SQL Server 2012
    47,500 0.97 EHP5 2014003
    IBM Power 760
    POWER7+ @3.41 GHz
    8 / 48 / 192 — 1024 GB
    AIX 7.1
    DB2 10
    25,488 0.99 EHP5 2013004

    Version – Version of SAP, EHP5 refers to SAP ERP 6.0 Enhancement Package 5 for SAP ERP 6.0 and EHP4 refers to SAP ERP 6.0 Enhancement Package 4 for SAP ERP 6.0

    Ch / Co / Th – Total chips, coreas and threads

    Complete benchmark results may be found at the SAP benchmark website http://www.sap.com/benchmark.

    Configuration Summary and Results

    Hardware Configuration:

    1 x SPARC M6-32 server with
    32 x 3.6 GHz SPARC M6 processors (total of 32 processors / 384 cores / 3072 threads)
    16 TB memory
    6 x Sun Server X3-2L each with
    2 x Intel Xeon E5-2609 2.4 GHz Processors
    16 GB Memory
    4 x Flash Accelerator F40
    12 x 3 TB SAS disks
    2 x Sun Server X3-2L each with
    2 x Intel Xeon E5-2609 2.4 GHz Processors
    16 GB Memory
    1 x 8-Port 6Gbps SAS-2 RAID PCI Express HBA
    12 x 3 TB SAS disks

    Software Configuration:

    Oracle Solaris 11
    SAP Enhancement Package 5 for SAP ERP 6.0
    Oracle Database 11g Release 2

    Certified Results (published by SAP)

    Number of SAP SD benchmark users:
    140,000
    Average dialog response time:
    0.58 seconds
    Throughput:

      Fully processed order line items per hour:
    15,878,670
      Dialog steps per hour:
    47,636,000
      SAPS:
    793,930
    Average database request time (dialog/update):
    0.020 sec / 0.041 sec
    SAP Certification:
    2014008

    Benchmark Description

    The SAP Standard Application SD (Sales and Distribution) Benchmark is an ERP business test that is indicative of full business workloads of complete order processing and invoice processing, and demonstrates the ability to run both the application and database software on a single system. The SAP Standard Application SD Benchmark represents the critical tasks performed in real-world ERP business environments.

    SAP is one of the premier world-wide ERP application providers, and maintains a suite of benchmark tests to demonstrate the performance of competitive systems on the various SAP products.

    See Also

    Disclosure Statement

    Two-tier SAP Sales and Distribution (SD) standard application benchmarks, SAP Enhancement Package 5 for SAP ERP 6.0 as of 3/26/14:

    SPARC M6-32 (32 processors, 384 cores, 3072 threads) 140,000 SAP SD users, 32 x 3.6 GHz SPARC M6, 16 TB memory, Oracle Database 11g, Oracle Solaris 11, Cert# 2014008. Fujitsu SPARC M10-4S (40 processors, 640 cores, 1280 threads) 153,000 SAP SD users, 40 x 3.0 GHz SPARC65 X, 10 TB memory, Oracle Database 11g, Oracle Solaris 11, Cert# 2013014. IBM Power 780 (12 processors, 96 cores, 384 threads) 57,024 SAP SD users, 12 x 3.72 GHz IBM POWER7+, 1536 GB memory, DB210, AIX7.1, Cert#2012033. Fujitsu PRIMEQUEST 2800E (8 processors, 120 cores, 240 threads) 47,500 SAP SD users, 8 x 2.8 GHz Intel Xeon Processor E7-8890 v2, 1024 GB memory, SQL Server 2012, Windows Server 2012 Standard Edition, Cert# 2014003. IBM Power 760 (8 processors, 48 cores, 192 threads) 25,488 SAP SD users, 8 x 3.41 GHz IBM POWER7+, 1024 GB memory, DB2 10, AIX 7.1, Cert#2013004.

    Two-tier SAP Sales and Distribution (SD) standard application benchmarks, SAP Enhancement Package 4 for SAP ERP 6.0 as of 3/26/14:

    IBM Power 795 (32 processors, 256 cores, 1024 threads) 126,063 SAP SD users, 32 x 4 GHz IBM POWER7, 4 TB memory, DB2 9.7, AIX7.1, Cert#2010046.

    SAP, R/3, reg TM of SAP AG in Germany and other countries. More info www.sap.com/benchmark

    About

    BestPerf is the source of Oracle performance expertise. In this blog, Oracle's Strategic Applications Engineering group explores Oracle's performance results and shares best practices learned from working on Enterprise-wide Applications.

    Index Pages
    Search

    Archives
    « July 2016
    SunMonTueWedThuFriSat
         
    1
    2
    3
    4
    5
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    26
    27
    28
    29
    30
    31
          
    Today