Hi everyone, i have this question for you i've always been curious about and now that i need to actually implement this it's time to get to an answer We have to run a pilot for 200 VDI u...
See more...
Hi everyone, i have this question for you i've always been curious about and now that i need to actually implement this it's time to get to an answer We have to run a pilot for 200 VDI users. We have our project all decided but one aspect is not so clear to me: how to meet IOPS requirement i have been given. For example: let's say that a consultant using the View Flash Calculator tells you that you need to support a total of 11200 frontend IOPS and a total of 20800 backend IOPS how would you provision your storage side? How many spindles and what type/rpm/technology? In other words, how do you do the math to know this in advance? For what i know most people recommends to consider VDI workload as 8K 80% writes and 20% reads 100% random. It would be easy to measure with Iometer if i already have the storage part, but what if i have to actually go and buy something? (a dedicate storage array or adding spindles to a present one) Also, how do you count in cache size? It really speeds things up but i can't tell in advance how many cache hits i will have so i figured i will have to consider the worst scenario and provision like the workload would hit the disks all the time. Thinking about this made me a little confused, leaving alone the frontend/backend storage IOPS part which is a little obscure to me by itself... How do you guys do this?