Why IOPS rocks
Today I got to know Benjamin's IOPS. It is a mini-disk-benchmark, so, comparable to iozone, bonnie, dd, dt and IOMeter.
It is much better than all of them. Here is why:
- every once in a while, I am asked "how many IOPS is your storage capable of?". With IOTest, I can answer this question, at least tell about one IOPS measurement. All other tools give me a multitude of values to choose from.
- bonnie is a great benchmark. But everytime when I start it, I ask first "what shall I write behind -e?". Ask whoever you want - the answer will be like "a big number, best, your RAM size". When I come back and tell the results, people normally say "this value was not big enough. You have to take at least twice your RAM size.". Everybody has his own opinion how big "big" is.
- IOTest does random reads. That means, it does not destroy your lun and it is not the usual sequential dd test.
Today I am using IOps again and I see some shortcomings:
- it is not clear to the user when IOps stops. For me I would always like to have some more values. Even analyzing the source code does not help quickly.