World Record Performance on PeopleSoft Enterprise Financials Benchmark run on Sun SPARC Enterprise M4000 and M5000
By Brian Whitney on Dec 02, 2010
Oracle's Sun SPARC Enterprise M4000 and M5000 servers have combined to produce a world record result on Oracle's PeopleSoft Enterprise Financial Management 9.0 benchmark.
The Sun SPARC Enterprise M4000 and M5000 servers configured with SPARC64 VII+ processors along with Oracle's Sun Storage F5100 Flash Array system achieved a world record result using PeopleSoft Enterprise Financial Management and Oracle Database 11g Release 2 software running on the Oracle Solaris 10 operating system.
The PeopleSoft Enterprise Financial Management 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 in record time.
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 Sun SPARC Enterprise M4000 and M5000 servers were able to process online users and concurrent batch jobs simultaneously in 34.72 minutes on this benchmark that reflects complex, multi-tier environment and utilizes a large back-end database of nearly 1 TB.
The combination of Oracle's PeopleSoft Enterprise Financial Management 9.00.00.331, PeopleSoft PeopleTools 8.49.23 and Oracle WebLogic server was run on the Sun SPARC Enterprise M4000 server and Oracle database 11g Release 2 was run on the Sun SPARC Enterprise M5000 server for this benchmark.
The following table discloses the current and the single previously disclosed result for this benchmark. Results are elapsed times therefore the smaller number is better.
|Sun SPARC Enterprise M4000||2.66 GHz SPARC64 VII+||Web/App|
|Sun SPARC Enterprise M5000||2.66 GHz SPARC64 VII+||DB|
|SPARC T3-1||1.65 GHz SPARC T3||Web/App||35.82||37.01|
|Sun SPARC Enterprise M5000||2.5 GHz SPARC64 VII||DB|
Web/Application Tier Configuration:
128 GB of memory
Database Tier Configuration:
128 GB of memory
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)
PeopleSoft Enterprise Financial Management/SCM 9.00.00.311 64-bit
PeopleSoft Enterprise (PeopleTools) 8.49.23 64-bit
Oracle Database 11g Release 2 18.104.22.168 64-bit
Oracle Tuxedo 9.1 RP36 with Jolt 9.1
Micro Focus COBOL Server Express 4.0 SP4 64-bit
This Day-in-the-Life benchmark measured the concurrent batch and online performance for a large database model. This scenario more accurately represents a production environment where users and scheduled batch jobs must run concurrently. This benchmark measured performance results during a Close-the-Books process.
The PeopleSoft Enterprise Financials 9 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, an 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
The Sun SPARC Enterprise M4000 and M5000 servers were able process online users and concurrent batch jobs simultaneously in 34.72 minutes.
The Sun Storage F5100 Flash Array system, which is highly tuned for IOPS, contributed to the result through reduced IO latency.
The family of Sun SPARC Enterprise M-series servers, with Sun Storage F5100 Flash Array systems, form an ideal environment for hosting complex multi-tier applications. This is the second public disclosure of any system running this benchmark.
The Sun SPARC Enterprise M4000 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.
Despite the systems being lightly loaded, the increased frequency of the SPARC64 VII+ processors yielded lower latencies and faster elapsed times than previously disclosed results.
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.
- PeopleSoft White Papers
- Dynamic SGA Tuning of Oracle Database on Oracle Solaris with DISM
- Deploying Oracle Database on the Oracle Solaris Platform – An Introduction
- Deploying Oracle Database on Oracle Sun SPARC Enterprise M-series Systems
Oracle's PeopleSoft Financials 9.0 benchmark, Oracle's Sun SPARC Enterprise M4000 (4 2.66 SPARC64 VII+), Oracle's Sun SPARC Enterprise M5000 (8 2.66 SPARC64 VII+), 34.72 min. Results as of 12/02/2010, see www.oracle.com/apps_benchmark/html/white-papers-peoplesoft.html for more about PeopleSoft.