SPARC T3-1 Performance on PeopleSoft Enterprise Financials 9.0 Benchmark
By Brian Whitney on Sep 23, 2010
Oracle's SPARC T3-1 and Sun SPARC Enterprise M5000 servers combined with Oracle's Sun Storage F5100 Flash Array storage has produced the first world-wide disclosure and World Record performance on the PeopleSoft Enterprise Financials 9.0 benchmark.
Using SPARC T3-1 and Sun SPARC Enterprise M5000 servers along with a Sun Storage F5100 Flash Array system, the Oracle solution processed online business transactions to support 1000 concurrent users using 32 application server threads with compliant response times while simultaneously completing complex batch jobs. This is the first publication of this benchmark by any vendor world-wide.
The Sun Storage F5100 Flash Array system is a high performance, high-density solid-state flash array which provides a read latency of only 0.5 msec which is about 10 times faster than the normal disk latencies of 5 msec measured on this benchmark.
The SPARC T3-1 and Sun SPARC Enterprise M5000 servers were able process online users and concurrent batch jobs simultaneously in 38.66 minutes on this benchmark that reflects complex, multi-tier environment and utilizes a large back-end database of nearly 1 TB.
Both the SPARC T3-1 and Sun SPARC Enterprise M5000 servers used the Oracle Solaris 10 operating system.
The combination of Oracle's PeopleSoft Enterprise Financials/SCM 9.00.00.331, PeopleSoft Enterprise (PeopleTools) 8.49.23 and Oracle WebLogic server running on the SPARC T3-1 server and the Oracle database 11g Release 1 was run on the Sun SPARC Enterprise M5000 server for this benchmark.
As the first world-wide disclosure of this benchmark, no competitive results exist with which the current result may be compared.
|Batch Processing Times|
|Batch Process||Elapsed Time in Minutes|
|Batch Alone\*||Batch with
1000 Online Users\*
|Total Main Batch Stream||20.69||26.66|
|Max SQR and nVision Stream||11.83||12.00|
|Total Batch (sum of Main Batch and Max SQR)||32.52||38.66|
\* PeopleSoft Enterprise Financials batch processing and post-processing elapsed times.
Results and Configuration Summary
1 x Sun SPARC Enterprise M5000 (8 x SPARC64 at 2.53 GHz with 64 GB of memory)
1 x Sun Storage F5100 Flash Array (74 x 24 GB FMODs)
2 x StorageTek 2540 (12 x 146 GB SAS 15K RPM)
1 x StorageTek 2501 (12 x 146 GB SAS 15K RPM)
1 x Dual-Port SAS Fibre Channel Host Bus Adapters (HBA)
Oracle's PeopleSoft Enterprise Financials/SCM 9.00.00.311 64-bit
Oracle's PeopleSoft Enterprise (PeopleTools) 8.49.23 64-bit
Oracle 11g R2 188.8.131.52 64-bit
Oracle Tuxedo 9.1 RP36 with Jolt 9.1
Micro Focus COBOL Server Express 4.0 SP4 64-bit
The PeopleSoft Enterprise Financials batch processes included in this benchmark are as follows:
Journal Generator: (AE) This process creates journals from accounting entries (AE) generated from various data sources, including non-PeopleSoft systems as well as PeopleSoft applications. In the benchmark, the Journal Generator (FS_JGEN) process is set up to create accounting entries from Oracle's PeopleSoft applications in the same database, such as PeopleSoft Enterprise Payables, Receivables, Asset Management, Expenses, Cash Management. The process is run with the option of Edit and Post turned on to edit and post the journals created by Journal generator. Journal Edit is an AE program and Post is a COBOL program.
Allocation: (AE) This process allocates balances held or accumulated in one or more entities to more than one business unit, department or other entities based on user-defined rules.
Journal Edit & Post: (AE & COBOL) Journal Edit validates journal transactions before posting them to the ledger. This validation ensures that journals are valid, for example: valid ChartFields values and combinations, debits and credits equal, and inter/intra-unit balanced, Journal Post process posts only valid, edited journals, ensures each journal line posts to the appropriate target detail ledgers, and then changes the journal's status to posted. In this benchmark, the Journal Edit & Post is also set up to edit and post Oracle's PeopleSoft applications from another database, such as PeopleSoft Enterprise Payroll data.
Summary Ledger: (AE) Summary Ledger processing summarizes detail ledger data across selected GL BUs. Summary Ledgers can be generated for reporting purposes or used in consolidations.
Consolidations: (COBOL) Consolidation processing summarizes ledger balances and generates elimination journal entries across business units based on user-defined rules.
SQR & nVision Reporting: Reporting will consist of nVision and SQR reports. A balance sheet, and income statement, and a trial balance will be generated for each GL BU by SQR processes GLS7002 and GLS7012. The consolidated results of the nVision reports are run by 10 nVision users using 4 standard delivered report request definitions such as BALANCE, INCOME, CONSBAL, and DEPTINC. Each of the nVision users will have ownership over 10 Business Units and each of the nVision users will submit multiple runs that are being executed in parallel to generate a total of 40 nVision reports.
Batch processes are run concurrently with more than 1000 emulated users executing 30 pre-defined online applications. Response times for the online applications are collected and must conform to a maximum time.
Key Points and Best Practices
Oracle's SPARC T3-1 and Oracle's Sun SPARC Enterprise M5000 servers published the first result for Oracle's PeopleSoft Enterprise Financials 9.0 benchmark for concurrent batch and 1000 online users using the large database model on Oracle 11g running Oracle Solaris 10.
The SPARC T3-1 and Sun SPARC Enterprise M5000 servers were able process online users and concurrent batch jobs simultaneously in 38.66 minutes.
The Sun Storage F5100 Flash Array system, which is highly tuned for IOPS, contributed to the result through reduced IO latency.
The combination of the SPARC T3-1 and Sun SPARC Enterprise M5000 servers, with a Sun Storage F5100 Flash Array system, form an ideal environment for hosting complex multi-tier applications. This is the first public disclosure of any system running this benchmark.
The SPARC T3-1 server hosted the web and application server tiers, providing good response time to emulated user requests. The benchmark specification allows 1000 users, but there is headroom for increased load.
The Sun SPARC Enterprise M5000 server was used for the database server along with a Sun Storage F5100 Flash Array system. The speed of the M-series server with the low latency of the Flash Array provided the overall low latency for user requests, even while completing complex batch jobs.
The parallelism of the SPARC T3-1 server, when used as an application and web server tier, is best taken advantage of by configuring sufficient server processes. With sufficient server processes distributed across the hardware cores, acceptable user response times are achieved.
The low latency of the Sun Storage F5100 Flash Array storage contributed to the excellent response times of emulated users by making data quickly available to the database back-end. The array was configured as several RAID 0 volumes and data was distributed across the volumes, maximizing storage bandwidth.
The transaction processing capacity of the Sun SPARC Enterprise M5000 server enabled very fast batch processing times while supporting over 1000 online users.
While running the maximum workload specified by the benchmark, the systems were lightly loaded, providing headroom to grow.
Please see the white paper for information on PeopleSoft payroll best practices using flash.
- Dynamic SGA Tuning of Oracle Database on Oracle Solaris with DISM (pdf)
- Deploying Oracle Database on the Oracle Solaris Platform – An Introduction (pdf)
- Deploying Oracle Database on Oracle Sun SPARC Enterprise M-series Systems (pdf)
Oracle's PeopleSoft Financials 9.0 benchmark, Oracle's SPARC T3-1 (1 1.65GHz SPARC-T3), Oracle's SPARC Enterprise M5000 (8 2.53GHz SPARC64), 38.66 min. www.oracle.com/apps_benchmark/html/white-papers-peoplesoft.html Results 09/20/2010.