Hi,
We have a 4.0f cluster attached to dual HSZ70s. Several of the services
we run on this cluster are very I/O intensive so we've been watching
disk activity for ways to improve performance. We see requests being
queued up on the server side. When we watch activity on the controllers
with vtdpy, however, the peaks (combining the rates on both controllers)
seem to be near 20M/sec rather than 40M/sec as expected.
Does anyone have suggestions on how to determine what is limiting I/O?
We've already verified that the cabling is correct and that the drives
support 40M/sec.
Thanks.
Teresa
Received on Fri Nov 08 2002 - 17:11:51 NZDT