Hi I am looking for what other admins and engineers look for when capacity planning. I have examples of items I look at, but would be interested in metrics used by others and how they verify those metrics.
Growth Rates of Volumes and Aggregates
Front End IOPs (how do I tell when I should be thinking about upgrading heads or adding additional NetApp hardware on this one)
Disk Utilization (If you use this metric what % of disk utilization do you like not to exceed. For example you want to add more disks to your aggregates before you reach and average utilization of x%?)
Latencies in volules/luns or do you look at array latencies for protocols or both
Thin Provisioning % over subscription or aggregate capacity (which metric do you look at do you look at both: For example you prefer not exceed more than x% over subscription and or your actual utilization within your aggregate is high with a high over subscription. Example again is 80% utilized capacity in aggregate with 200% over subscription is bad (or maybe not depending on your growth rates)
Any idea and or metrics others use would be very helpful and data to go along with how you use those metrics would be even more useful.
I am looking for what other storage admins and or architects do for capacity planning of existing arrays. At some point all storage admins already or should be looking at there existing storage systems to see when they need to order additional capacity or add new or upgraded hardware to handle growing workloads or new workloads.
I am looking for the general ways that current admins do this with there arrays now. I am finding it hard to believe that no one does this today.
We have a lot of customers with a shared storage. When it comes to plan a upgrade, we take a look at the hardware which is already in use. In most of our cases it is really easy to find a new storage to replace it with the old one.
I can give you an easy example:
A customer uses 4 esx hosts with a netapp 2040 and a additonal shelf, in summary with 36 x 300 SAS 15k 3,5" in an FC-environment and has 50 VMs running. Let's face it, if the customer has any performance issues, he would let you know about it and asks how to get rid of it. Then you have to analyse and find a way to get better performance. But let's pretend, the customer is quite satisfied with his environment. When it comes to replace the storage, and there is a normal growth of the infrastructure I would recommend a FAS2240 and at least one additional shelf with summary 48 x 600 SAS 10k 2,5". If the budget is big enough, you can add an second shelf with 24 disks.
I don't know in which sizes you think, but in a small environment (like my example) is this a way to go for me. We did this a lot of times and never had a problem with storage decisions. But I have to say, we know our customers for years and have a good overview of thier IT.