JVM Performance

[Anthony Leonard]

SPECjvm98 scores from www.spec.org:

B2000, 1x 400MHz CPU = 47.4
A500, 2x 440MHz CPUs = 54.1
C3600, 1x 552MHz CPU = 71.5

So the A500 seems only marginally faster than the B2000, even equipped with 2 faster CPUs - why?

My guess is that the SPECjvm98 benchmarks suite doesn't make full use of multiple processors - so the A500 result is almost equivalent to an A500 with 1x 440MHz CPU. Then the numbers make more sense.

However this assumes the C-class doesn't have anything else in it to make it's JVM faster other than the 552MHz CPU. Is that true?

Any insights appreciated...
(I'm hoping that, if I'm right, the A500 with 2x 552 MHz CPUs will be really fast for an app that scales across processors....)

[Keith Hughes_2]

Hi,

Have you used the HPjconfig tool.

Please find the download at

http://www.unix.hp.com/java/java2/hpjconfig/index.html.

This tool will be very helpfull in giving you the optimum setting of your kernel parameters. On setting them, the performance of your different systems should inprove as the UNIX box has been setup to run your Java application.

This may well improve the performance of your multi-processor system. The performance of the systems may also change by small changes to the Java application.

Happy Coding

Keith