Hi JNV,
It's a bit "how long is a piece of string?" question...
What's the size of the IO? Is it random or sequential? What latency is acceptable?
As a rule of thumb, for short (4K or 8K) IO & reasonable latency (sub-20ms), each 'active' 15k spindle should deliver somewhere in the range of 180-220 IOPS.
'Active' spindles are your data drives, but not parity & hot spares, so as an example you may have following allocation (per each controller):
- 8 x data + 2x parity + 2x hot-spare
Regards,
Radek