disk throughput as seen through MW or Glance

[Deborah McIntosh]

As seen through Measureware, 14 meg/sec phys read is the maximum throughput we get, regardless of our disk sub-system. We had been using a VA7100 and recently switched to an EMC/Clarion CX400. They are connected via A5158A FC cards. Specs on both systems claim their throughput will be higher than this, so I don't know where this limitation is coming from. We have vxfs filesystems defined, and are running an Oracle 8 database. Using sar -b , the %rcache throughout the day can often be 0, which doesn't surprise me considering the amount of physical reads that happen on the database files. On the db side, the buffer cache hit ration is around 95% which is good. We are running PeopleSoft 7.5 applications. Any clues as to where this limitation is coming from? Or am I misinterpreting the statistics?

[B. Hulst]

Hi,

Could be an application bottleneck but try this...

Run sar or mw on the system to measure the activity.

Then do a read with dd on a very big device.

dd if=/dev/ of=/dev/null

This basically bypasses the applications, like oracle and give a bigger number of what the hardware should be able to handle...

Regards,
Bob

[Deborah McIntosh]

thanks - I will try this during a non-peak hour, since it is our db production server.

[B. Hulst]

Hi,

Any luck on this? Do you get some raw data from the dd?

Regards,
Bob