I want to know if the IOPs resource tuning is true physical IO operations ?
By this i mean that a series of sequential reads can cause a very high logical read rate while the physical disk accesses can be much lower due to cache buffers at various parts of the system.
e.g. two different senarios
a - sequential reads of some flat file can occur if allowed to run free in the 10s to 100s of Ks, lets say 100 logical to 1 physical
b - random reads or writes that require activity on the disk, lets say 1 for 1
What i am concerned about is if i clamp a vm to have 500 IOPs say, will case a) run slowly ?
The backdrop is we occasionally get a vm sucking all the IO out of our SAN which is shared by vCenter, Lab Manager and vCD, because by default these have "unlimited" for IOPs
p.s. interested in an KB / whitepaper which covers this in detail