If you are using standard Oracle tools for checking I/O response
times, then I can tell you that they match VM Toolkit reporting.
When we ran on aix/sun and compared to zVM and Linux on z-Series, we
found the numbers to be very much the same.   The only thing that the
Oracle tools cannot do is the CPU (SLES 9.x and before).   This is a
kernel issue that is mostly corrected in 10.
What really matters is the response time for all of the I/O's.   Quite
often the bottleneck is in the spreading of the data so no one single
device slows down everybody else.   Care should be taken in making
sure Oracle logs go as fast as possible.  Other than that, if your
paging/swapping is only going to VIO (expanded most likely on zVM),
then the I/O's should be the same and the comparison becomes one of
CPU and architecture.   Under z-Series architecture, I/O's are handled
by parallel processing on the I/O CPU's.   In the other distributed
system architectures, I/O's are by sar as CPU usage to support I/O's.
I found that when we took a heavy I/O bound application under AIX
running at 30% CPU just for I/O, we saved all of those cycles on the
z9.
By the way, we also saw signficant improvements in batch processing.
Jobs that were taking from 7 to 15 hours to run, were all of a sudden
running in 3 hours or less.

--
Yours truly,
Yu

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to