> I ran some tests with the 'dbench' benchmark from the samba project which
> simulates the disk i/o of the 'netbench' benchmark.

Make sure to know what you measure. With z/VM it may depend a bit... ;-)

Dunno about dbench, but I do know that some of these test programs from
another universe tend to measure time or timing by CPU loops and then
implicitly assume a relation between userland time and wall clock time.
Yes, I know this would at best shift the numbers the other way. However,
when I ran 'iperf' and 'netperf' benchmarks I found that the timing issue did
confuse the programs in such a way that they computed a negative delay
between transactions and things like that.

Rob

Reply via email to