The Oracle BI EE workload was run on two Sun SPARC Enterprise T5440 serversand acheived world record performance.
  • Two Sun SPARC Enterprise T5440 servers with four 1.6 GHz UltraSPARC T2Plus processors delivered the best performance of 50K concurrent userson the Oracle BI EE benchmark with Oracle 11g database runningon free and open Solaris 10.

  • The two node Sun SPARC Enterprise T5440 servers with Oracle BI EE running onSolaris 10 using 8 Solaris Containers shows 1.8x scaling over Sun'sprevious one node SPARC Enterprise T5440 server result with 4 SolarisContainers.

  • The two node SPARC Enterprise T5440 servers demonstrated the performance andscalability of the UltraSPARC T2 Plus processor demonstrating 50K users can be servicedwith 0.2776 sec response time.

  • The Sun SPARC Enterprise T5220 server was used as an NFS server with 4internal SSDs and the ZFS file system which showed significant I/Operformance improvement over traditional disk for Business IntelligenceWeb Catalog activity.

  • IBM has not published any POWER6 processor based results on this important benchmark.

Performance Landscape

System Processors Users Chips GHz Type 2 x Sun SPARC Enterprise T5440 8 1.6 UltraSPARC T2 Plus 50,000 1 x Sun SPARC Enterprise T5440 4 1.6 UltraSPARC T2 Plus 28,000 5 x Sun Fire T2000 1 1.2 UltraSPARC T1 10,000 Results and Configuration Summary

Hardware Configuration:

  • 2 x Sun SPARC Enterprise T5440 (1.6GHz/128GB)
    1 x Sun SPARC Enterprise T5220 (1.2GHz/64GB) and 4 SSDs (used as NFS server)
Software Configuration:

  • Solaris10 05/09
    Oracle BI EE
    Oracle 11gR1
Benchmark Description

The objective of this benchmark is to highlight how Oracle BI EE cansupport pervasive deployments in large enterprises, using minimalhardware, by simulating an organization that needs to support more than25,000 active concurrent users, each operating in mixed mode: ad-hocreporting, application development, and report viewing. The user population was divided into a mix of administrative users andbusiness users. A maximum of 28,000 concurrent users were activelyinteracting and working in the system during the steady-state period.The tests executed 580 transactions per second, with think times of 60seconds per user, between requests. In the test scenario 95% of theworkload consisted of business users viewing reports and navigatingwithin dashboards. The remaining 5% of the concurrent users,categorized as administrative users, were doing applicationdevelopment.

The benchmark scenario used a typical business user sequence ofdashboard navigation, report viewing, and drill down. For example, aService Manager logs into the system and navigates to his own set ofdashboards viz. .Service Manager.. The user then selects the .ServiceEffectiveness. dashboard, which shows him four distinct reports,.Service Request Trend., .First Time Fix Rate., .Activity ProblemAreas., and .Cost Per completed Service Call . 2002 till 2005. . Theuser then proceeds to view the .Customer Satisfaction. dashboard, whichalso contains a set of 4 related reports. He then proceeds todrill-down on some of the reports to see the detail data. Then the userproceeds to more dashboards, for example .Customer Satisfaction. and.Service Request Overview.. After navigating through these dashboards,he logs out of the application

This benchmark did not use a synthetic database schema. The benchmarktests were run on a full production version of the Oracle BusinessIntelligence Applications with a fully populated underlying databaseschema. The business processes in the test scenario closely representsa true customer scenario.

See Also

Disclosure Statement

Oracle BI EE benchmark results 10/13/2009, see