One of the papers that has haunted me ever since I first read it in college is called "The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information" (George A. Miller, 1956). It begins “My problem is that I have been persecuted by an integer” and proves that the number of objects an average human can hold in working memory is 7 ± 2.
This week, Oracle seems to be violating this very limit by overloading us with information; they published 30 or so press releases alone yesterday. With all this information overload, how can you make sure not to get distracted and confused? Here are the key performance points that you need to take away from all the recent OpenWorld benchmark results:
Watch the benchmarks used. Oracle compares SAP SD Parallel benchmark data with SAP SD standard data. Note that these benchmarks are different benchmarks and have different data distributions. Watch all the components of the benchmark. The SPECjAppServer2004 benchmark results only focus on the application server and not the key database server. Remember that when you are doing any compares. Oh, and these results are not even run on Oracle/Sun systems. Watch the configurations used in the benchmarks carefully. A SPECjEnterprise2010 IBM POWER7 result with WebSphere and DB2 was actually over 1.5 times the performance per core of a new Oracle SPARC T3 server result.(1) Watch for "internal benchmarks." Most of the new Oracle SPARC T3 benchmarks announced were actually Oracle's own applications - Siebel, Oracle, PeopleSoft, JD Edwards, even something called an "in-house workload." Watch for no benchmarks. The Oracle Exalogic Elastic Cloud and Exadata X2-8 systems announced have many claims but no real industry standard performance data that we can compare to. Just like the old Exadata V2, where we've been waiting over a year for any benchmark.
Note that I used 7-2=5 points above. Because who even remembers a 7 digit phone number these days.
(1)SPECjEnterprise2010 Oracle WebLogic Server 11g on a Oracle SPARC T3-4 server (64-core) and Oracle Database 11g on SPARCT3-2 server (32-core) , 9,456.28 SPECjEnterprise2010 EjOPS vs. WebSphere Application Server V7 on IBM Power 750 Express (32-core) and DB2 9.7 on IBM BladeCenter PS702 Express (16-core), 7,172.93 SPECjEnterprise2010 EjOPS.
Source: www.spec.org. Results current as of 9/21/10.
SPEC, SPECint, SPECfp, SPECjbb, SPECweb, SPECjAppServer, SPECjEnterprise, SPECvirt, SPECompM, SPECompL, SPECsfs, SPECpower, SPEC MPI and SPECpower_ssj are trademarks of the Standard Performance Evaluation Corporation (SPEC).
SAP, mySAP and other SAP product and service names mentioned herein as well as their respective logos are trademarks or registered trademarks of SAP AG in Germany and in several other countries all over the world.
The postings on this site solely reflect the personal views of the author and do not necessarily represent the views, positions, strategies or opinions of IBM or IBM management.
technorati tags: IBM, systems, performance, SPECjAppServer2004, SPECjEnterprise2010,750,SPARC,T3,WebSphere,PS702,benchmark,Oracle,OpenWorld,Exalogic,cloud,Java,HP,POWER7,DL980,DB2,Exadata,Weblogic,benchmarking