Benchmarking Applications Performance

It's not highly visible, but we have a team within the Applications Technology Group that's dedicated to optimizing the performance of the E-Business Suite.

Benchmark Graph:

This group covers it all:  runtime performance analysis for the database, application and client tiers, as well as performance analysis of our own internal development environments.  At any given time, they can be found analyzing JVM garbage collection parameters, JDBC connection pool logs, network roundtrips and latency, process address space allocations, and other related esoterica.

Interesting trivia: they maintain a comprehensive SQL repository with over 500,000 SQL statements used in a given E-Business Suite release, extracted directly from the source code, and the execution plans of these SQL statements based on volume-testing databases.  E-Business Suite developers use this database to identify potential performance issues with their specific apps products.

Our performance team also works with selected vendors to create standardized (and independently audited) Applications Benchmarks.  As of today, benchmarks are available for 11.5.9 and 11.5.10.  Interesting reading.

Note:  The graph shown above is for the Oracle e-Business Suite eBS Benchmark 11.5.10 using Oracle10g on HP ProLiant and Blade servers.


Hi Steven, How does the team load test a system? We have been looking into products to load test our instance before go-live and were shocked by the price of Load Runner but have not found any less expensive products that work with OAF pages.


Posted by Mark Harrington on May 31, 2006 at 11:52 PM PDT #

Hi, Mark,A timely question... I'd been thinking about writing about this for a while.  In response to your suggestion, I've just posted this:Automated Testing for the E-Business SuiteI'm not personally familiar with Mercury's pricing terms, but it doesn't surprise me to hear your reaction.  If I were in a position to evaluate the cost-benefit equation, I'd think have to look carefully at how much testing time the tool can save you.  The more end-users you have, the more power they have to veto new patches.  So, a number of our enterprise (read: LARGE) customers see this investment in strategic terms:  they enlist the help of their end-user community in building the tests initially, which reduces the anxiety and resistance around applying new patches.  Regards,Steven

Posted by Steven Chan on June 01, 2006 at 06:49 AM PDT #

broken link :

Posted by guest on July 02, 2013 at 09:37 AM PDT #

Thanks. Fixed now:


Posted by Steven Chan on July 02, 2013 at 09:58 AM PDT #

Thanks Steven.

The link " Oracle e-Business Suite eBS Benchmark 11.5.10 using Oracle10g on HP ProLiant and Blade servers." is broken, and its hard to get the full take away of this article without this information. Perhaps you could include it in the text of the article itself (primarily the CPU - type and numbers, RAM, virtual memory).

Another observation - If these are on a single server, the average CPU utilizations add up to 100% or more. This in itself is not a problem. But what would help is a point of time snapshot of the total CPU utilization across all 3. From my memory ATG would track that.

I would further want to make an analysis of per 100 users added, how does the load on the CPU and RAM increase - thats very useful information.


Posted by guest on August 22, 2013 at 02:18 AM PDT #

Hi, Prashanth,

Link fixed; you're welcome to review the detailed benchmark results directly in the report.

All of the published benchmarks show incremental loads. You can select a benchmark that most closely matches your hardware setup and the EBS modules that you're using.


Posted by Steven Chan on August 22, 2013 at 09:05 AM PDT #

Post a Comment:
  • HTML Syntax: NOT allowed


« November 2015