The 10k drives are about 30% slower for random IO than the 15k drives.

And since you are going to drives twice the size, you will need half as many for the same capacity.

So if your current drives have very low utilization, you should be OK to move to the 10k, but the response times for read miss IO will always be higher on the 10k than the 15k, even at low utilization.

Tclark,

This will be a classic case of Application IO profiling to make that determination. You will need to look at the following components:
- what is the max (and maybe 95th percentile) IOPS and MBPS numbers for the server?
- Are there any IO bottlenecks in the current IO subsystem in terms of device queues and disk utilization?
- what are the current service times and will they be negatively impacted due to disk RPM being slower? if the current disk utilization is low then it may not matter as long as you can provide as many spindle in the new env.
- Are you planning to keep the same RAID levels? the IO profile will show if you can change from RAID10 to RAID-5 etc
- You should start with using a OS tool (iostat, perfmon etc) to see what the IO profiles look like
- Also look at the array IO profile for the devices being used by this application.
- These steps can provide a good start. Modeling tool can help further (I have created few for Symmetrix and CLARiIONs)

I have done this for many arrays at a time to reclaim storage. sometimes you can be surprised on how much little IO is going on with many of the applications. Also it is not necessary that 100% of the app live on same type of disks and RAID levels (though it introduces some complexity)
Hope this helps...

Thank you both. We did move to the larger drives with no performance problems so far. Not that I expect any.