Short background: We have inherited a rather messy setup with ~30 ESX hosts and ~150 datastores with almost 500 guest VMs. Unfortunately the naming and organization is very chaotic.
We are looking into purchasing new storage systems and I would like to get a baseline IOPS requirement across our virtualized environment while we begin searching for solutions. This environment includes a mid-size View deployment but I have a good grasp on the I/O requirements for our VDI.
Is there any way to collect data such as average, peak, etc IOPS across all of our ESX hosts over a period of several days or maybe even a week? I can view the CMD/s in esxtop or read/write operations under performance for the storage adapter on each individual esx host but would like to avoid monitoring each host then combining the results. I'm not sure a flat average is really what I want either since it is rather low but I see fairly high spikes regularly.
Are there any scripts or tools available to monitor CMD/s or similar stats across multiple ESX hosts? Also, any advice, best practices, blogs, suggestions etc about evaluating storage I/O requirements across a rather large infrastructure? I feel that adding the average CMD/s per host together will not give me a realistic I/O requirement. Any help would be appreciated!