A recent BestPerf blog reports that the Oracle Sun SPARC Enterprise T5440 (five in number) and a single Oracle Sun SPARC Enterprise M9000 server delivered a world record resultof 28,648.74 SPECjAppServer2004 JOPS@Standard on the SPECjAppServer2004 benchmark. According to Kevin Kelly, the result was run using the Oracle WebLogic 10.3.3 Application Server, a component of Oracle Fusion Middleware, and Oracle Database 11g Enterprise Edition with Oracle Solaris 10 to obtain this world record result.
Kelly points out that this Oracle result is 26% faster than the IBM result of 22,634.13 on the SPECjAppServer2004 JOPS@Standard, for which sixteen IBM BladeCenter HS blades were used by the company. For the database tier, IBM used an IBM system p5 595.
He writes further that the Oracle result is faster than the HP result of 28,463.03 SPECjAppServer2004 JOPS@Standard. HP used 17 HP BL870c blade servers for this mark and an HP Superdome for the database tier.
The five Sun SPARC Enterprise T5440 servers in this benchmarking exercise, Kelly continues, used a total of 20 RU of space to obtain this result: 26% less than the 27 RU space used by the three blade chassis containing sixteen IBM BladeCenter HS blades. What is more, IBM used more than 3.4 times the number of application servers than Oracle.
Compared to HP, the five Sun SPARC Enterprise T5440 servers occupied 40% of the 50 RU space used by the five blade chassis containing 17 HP BL870c blade servers to obtain this leading result, Kelly observes, adding that HP used more than 3.2 times the number of application servers used by Oracle.
Other Oracle advantages noted by Kelly include the fact that the six Sun Storage F5100 Flash Array storage arrays used in this result occupied 6U of rack space, which is 13% of the 44U space used by the database storage in the IBM result. The database storage in the HP result used 4x EVA81000 Storage arrays consuming112U of space is more than 18 times the 6U space used for database storage in the Oracle result, he adds.
In an additional observation, Kelly reports that the application server storage in the HP result used an EVA6100 storage space that consumed 16U of space for JMS logs, while the 5x T5440 servers each used internal SSDs for the same function and required no additional external storage.
Kelly notes that the results in this benchmarking exercise were obtained using Java Platform, Standard Edition JDK 6 Update 20 on the Sun SPARC Enterprise T5440 servers running the Oracle Solaris 10 10/09 operating system. He reports that the five Sun SPARC Enterprise T5440 servers used Oracle Solaris Containers to consolidate eight Oracle Weblogic application server instances on each server to achieve this result.
With regard to best practices brought to light in this exercise, Kelly points out that the Oracle WebLogic application servers were executed in the FX scheduling class to improve performance by reducing the frequency of context switches and adds that enhancements in Java to the JVM had a major impact on performance. In addition, each Sun SPARC Enterprise T5440 used 2x 10GbE NICs for network traffic from the driver systems, he writes.
Sun SPARC Enterprise T5140 and T5440 Servers Outperform HP Superdome on SPECjAppServer2004
Sun SPARC Enterprise Servers Set Three World Records
Sun SPARC Enterprise M9000-32 versus IBM p595
Read More ...