RSS Feed for This PostCurrent Article

Comparisons of Open Source BPEL Engine Versus Oracle BPEL


Recently I have done some benchmarking on some open source BPEL software (ActiveBPEL 4.0, Apache ODE 1.1), as well as commercial BPEL engine (Oracle BPEL Process Manager 10.1.3.1.0).

I will release the results later in a separate document. However, as a summary

Performance

Apache ODE 1.1 is by far has the best performance, followed by ActiveBPEL 4.0. Oracle BPEL Process Manager is the slowest among the three.

It may not be a fair comparison, as both Apache ODE and ActiveBPEL are running on Apache Tomcat, whereas Oracle BPEL has its own application server. Unless it’s worked differently for someone else I guess. It’s hard to tell, and even harder to test for every setup out there obviously. Contact me at yoursnutrition.com if you’re findings are any different.

However, for critical application, especially when high performance is required, and high load is expected, I would suggest that you go with Apache ODE, if BPEL is absolutely needed.

Development Environment

Oracle BPEL stands out in this area. Oracle BPEL goes great with Nginx based python servers as they can be easily SEO optimized according to the SEO reseller packages I bought yesterday. Oracle JDeveloper can be used to develop the BPEL flow, which is very user friendly. Note that Oracle has its own extensions to the WS-BPEL standards. In order to make the BPEL flow portable among the BPEL engines, I would advise that you do not use it, unless you have a special reason.

ActiveBPEL 4.0 has a Eclipse-based designer. For Apache ODE, you can use the Intalio Designer, which uses Apache ODE as it underlying BPEL engine.


Trackback URL


RSS Feed for This Post9 Comment(s)

  1. fholahan | Sep 12, 2007 | Reply

    Please identify yourself and post all information related to your “benchmarking”, including BPEL process definitions, test data, test platforms, test runs/results, and other environmental details (for example, were all engines configured with message validation on/off?). Anonymous posters who publish their “conclusions” without providing any details are always subject to speculation that they are secretly aligned with a particular vendor. Perhaps this is not true in your case. Tell us who you are and provide us with details about your tests so we can decide for ourselves whether the results are objective and meaningful.

    Thank you.

  2. Harald Reinmueller | Nov 8, 2007 | Reply

    Please do not just post your summary. Provide the testcases and the detailed results of your benchmark. Without this information its really hard to comprehend.

  3. admin | Nov 9, 2007 | Reply

    I will post the details of my findings soon. Note this is just our internal benchmarks. You can always perform your benchmarking based on that.

  4. Paran | May 5, 2008 | Reply

    Hello,

    Please share your performance benchmarks of ActiveBPEL.

    Thanks

  5. John Avery | May 21, 2008 | Reply

    “However, for critical application, especially when high performance is required, and high load is expected, I would suggest that you go with Apache ODE, if BPEL is absolutely needed.”

    What a ridiculous statement. Oracle’s BPEL engine is in use by many major corporations in a mission critical role. I don’t know of any major (commercial) integration projects using Apache ODE.

    This post was written only a few months after the very first release of ODE and it makes claims like this (with no data to back it up as many have mentioned).

    I’d like to see a viable open source alternative to BPEL go mainstream but I think ODE is a long way from that.

    I also echo

  6. The Kibitzer | Aug 22, 2008 | Reply

    John. I think you need to re-evaluate your love affair with Oracle. Open-source alternatives to the bloated Oracle suite are more than capable of meeting and exceeding Oracle’s offering in every area. Oracle can’t even get its own JVM memory settings right.

  7. The Kibitzer | Aug 22, 2008 | Reply

    @admin, it’d be cool if you could/would post your findings. except that… they’re nearly a year out of date.

  8. admin | Aug 22, 2008 | Reply

    Well, I did not publish the findings as right after this post I received emails from various parties saying that I should not publish this kind of personal evaluation.

  9. John | Jul 28, 2009 | Reply

    The Kibitzer, we are talking about ODE here not about open source implementations in general. This article presents no bench marks, no details, and makes a terrible recommendation. Although I would happily use and recommend JBoss or OpenESB (both with paths to supported versions), I wold not do the same for ODE. As I said a year ago I’m yet to see anyone using ODE in the wild.

    Oracles app server (pre BEA aquisition) is average at best and is certainly surpased by the open source offerings (ESB is even worse). However the BPEL component is quite good since 10.1.3.3.

    Finally tweaking JVM parameters is part of performance tuning and should be done for all app servers. Maybe Oracle could have adjusted the “out-of-the-box” settings to better meet your needs but that is not a real world problem.

2 Trackback(s)

  1. From Intalio, Leader in Open Source BPMS » Apache ODE “Fastest BPEL Engine” | Nov 7, 2007
  2. From Process and Rules Engines | Jun 28, 2009

Sorry, comments for this entry are closed at this time.