I understand cumulating IOs under datastore - Disk I/O makes sense to have a single view of how many IOs are thrown from a specific host/VM or against a specific datastore.
But when it comes to latencies, this makes no sense at all.
For example, monitoring READ Latency on all datastores from a VM sight shows peaks up to 250ms - which could mean every VM has a latency of 0,25ms (as we utilize about 1000VMs) or no VMs show latencies except one, which has 250ms.
Is there an ability to change this setting for latencies?
If not, could this be realized in coming versions?