I don't know if this is even possible to gather granular (< 20 sec.) stats via VMware APIs, but it would be nice if Veeam ONE were able to enable an on demand real time monitoring for let’s say 24 hrs.
I spent some time to compare I/O stats gathered by perfmon (inside the VM), esxtop, vCenter performance graphs and on the storage backend.
As expected the values gathered by esxtop (2 sec. interval) and the storage backend were close to be identical.
The vCenter performance graphs sometimes reported 50% lower values as esxtop/the backend.
This is due to the fact that the graph uses 20 sec intervals and all collected values in between will be “averaged”.
So even if there is a low base load < 20 I/Os, a high peek let’s say 2000 I/Os, and an average of 500 I/Os, the vCenter graph will report ~ 1000 I/Os for the last 20 sec. And if you rely only on these stats you may be far of what’s really happening.
As non-native speaker it’s not easy to explain such a complicated scenario, sorry for that
So If there’s a way to gather more detailed stats than the default 20 sec. it would be awesome to implement a kind of on demand real time monitoring.
Let’s say this mode would enable me to set/select:
Duration: like 24 hrs.
Stats: CPU/MEM/Disk(I/O) to reduce the performance impact
This would offer a detailed overview of what’s going on and will help me to plan future extensions, based on reliable stats.