Why WebSphere? Blog

Last few weeks brought us two new SPECjEnterprise2010 results – one from Oracle and one from IBM. Both were done using very latest software and hardware. Oracle announced their new SPARC T5 processor with much fanfare and claiming it to be the “fastest processor in the world”. Well, perhaps it is the fastest processor that Oracle has produced, but certainly not the fastest in the world. You see, when you publish industry benchmarks, people may actually compare your results to other vendor’s results. This is exactly what I would like to do in this article.

Full results can be found here:
Oracle total EjOPS: 57,422.17 http://www.spec.org/jEnterprise2010/results/res2013q1/jEnterprise2010-20130305-00041.html
IBM total EjOPS: 13,161.07 http://www.spec.org/jEnterprise2010/results/res2013q2/jEnterprise2010-20130402-00042.html

Being “fastest processor in the world” means that such processor must be able to handle the most transactions per second per processor core, which is how software pricing works and how people size their workloads and control their costs. This is not the proof Oracle delivered with their latest result (see full details on Spec website). To give Oracle credit, their result is the biggest overall 57,422.17 EjOPS (transactions per second). But that is a Total number of transactions, not a measure of the processor speed. To achieve that result, Oracle had to use 128 SPARC T5 cores for the WebLogic 12c and additional 128 cores for the Oracle database! The total cost of the hardware to achieve such high number of Total EjOPS is $1.1 Million. Even more sobering is the list price for the software, which is $5.2 Million (including 3 years of support and using lower priced WebLogic Standard – not even clustered!). If you price Oracle configuration with the WebLogic Enterprise (which does support clustering), your software cost will be $7.7 Million. Overall this latest Oracle result produced 449 EjOPS/core at the cost of $109.45 per EjOPS.

Now look at the IBM result published recently using WebSphere 8.5.5 on Power7+ hardware with DB2 database. IBM did not go after the biggest number of EjOPS (which is just the matter of throwing bunch of hardware together). However IBM produced the world record result in terms of EjOPS per processor core – truly a measure of the fastest processor known to men (for Java EE workloads that is). The total hardware cost of IBM result is $74,000 and the software cost is $766,000 (of which WebSphere is only $72,000 and the rest is DB2). This IBM result delivered world record 823 EjOPS per core with the cost of $63.79 per EjOPS. Now this is almost twice as many transactions per second at almost half of the Oracle cost. Truly remarkable.

Since Oracle knew they can not produce the most efficient result in terms of cost or transactions per second, the only way for them to claim world record was to throw large hardware at it and produce the biggest total number of EjOPS. Not a very useful metric I must admit. Much more interesting is the efficiency – measured in EjOPS per core and most importantly cost of EjOPS.

The story does not end here. Why not take a look at the history of performance results on similar and dissimilar hardware? Why not compare these platforms:

• IBM WebSphere on Power7+ vs. Oracle WebLogic on SPARC T5 (latest generation hardware – shown above, but just to rub it in)
• IBM WebSphere on Power7 vs. Oracle WebLogic on SPARC T4 (previous generation hardware for both vendors)
• IBM WebSphere vs. Oracle WebLogic on Intel Sandy Bridge Xeon E5-2690 (almost identical hardware setup using latest Intel hardware)
• IBM WebSphere vs. Oracle WebLogic on Intel Westmere Xeon X5690 (almost identical hardware setup using older Intel hardware)

Here is a summary of these results listed above:

Here is a brief summary of the IBM WebSphere performance track record since year 2000:

• Held the most records in ECPerf (pre-2001)
• FIRST to publish SPECj2001
• FIRST to publish SPECj2002
• FIRST and ONLY company to publish SPECj2002 Distributed
• FIRST to publish SPECj2004 and the only vendor to publish for over 13 months, held #1 spot for most of the time
• FIRST to publish SPECjEnterprise2010
  • LOWEST cost per transaction as of today
  • BEST performance per core as of today

For additional information, please refer to these performance related articles: http://whywebsphere.com/?s=specj

******************* Notes:

(1) SPEC and SPECjEnterprise2010 are registered trademarks of the Standard Performance Evaluation Corporation. Results from http://www.spec.org as of 04/04/2013 Oracle SUN SPARC T5-8 449 EjOPS/core SPECjEnterprise2010 (Oracle’s WLS best SPECjEnterprise2010 EjOPS/core result on SPARC). IBM Power730 823 EjOPS/core (World Record SPECjEnterprise2010 EJOPS/core result), (2) Results from http://www.spec.org as of 04/29/2012 Oracle SUN SPARC T4-4 313 EjOPS/core SPECjEnterprise2010 (Oracle’s WLS best SPECjEnterprise2010 EjOPS/core result on SPARC). IBM Power780 681 EjOPS/core (World Record SPECjEnterprise2010 EJOPS/core result), (3) Results from http://www.spec.org as of 11/14/2012 Oracle SUN Fire X4170M3 519.39 EjOPS/core SPECjEnterprise2010 (Oracle’s WLS best SPECjEnterprise2010 EjOPS/core result on Sandy Bridge). IBM WAS 8.5 System x3650 M4 Intel Sandy Bridge EjOPS/core (World Record SPECjEnterprise2010 EJOPS/core result) (4) Results from http://www.spec.org as of 04/29/2012 Oracle SUN Blade Server X6270 M2 452.285 EjOPS/core SPECjEnterprise2010). IBM Websphere HS 22 Blade 524.621 EjOPS/core.