The Truth on IOPS in Storage Shopping
When we are looking at criteria for making a decision on storage, one of the top items always seems to be IOPS (I/O Per Second). As we do more and more work around storage performance measurement, this number is always featured, but is it the right way to make your hardware decision?
IOPS Are Important, But Are They Always the Same?
You may be thinking that IOPS are always IOPS. Let’s say that as an example that we know the storage workload we are running requires 87,250 IOPS, so we can size accordingly. However, there should be much more in the decision process as we profile the behavior of our workloads.
It is possible to use many different storage configurations to achieve the same number of IOPS, but there are other factors that are affected including total storage available, latency, and physical topology design that may not account for redundancy and resiliency.
Melissa Palmer did a great article called Stick a Needle in Your IOPS that looks at just such a storage decision that shows an interesting view of the situation.