While involved in some storage latency analysis on vCenter 6.0 / ESXi 6.0 recently, and I noticed that the "Realtime - Average Read Latency - (Top 10)" graph in the Web Client (Figure #1 below) seems to be reporting average latencies about 100x higher than what we are seeing in the "Realtime - VM Average Read Latency" graph in the vSphere Client (Figure #2 below). I noticed this at a customer site as well as in our lab environment (both running vSphere 6.0). Is this perhaps a bug in the Web Client, or am I interpreting the graphs incorrectly?
In the example below, Figure #1 shows average latency around 200ms for a VM around 8:45am, while Figure #2 shows average latency under 1ms over the last hour for the same VM, with a few spikes to 2-3ms around 8:45am.
The blue colors are hard to distinguish in Figure #1, so I checked all of the VMs that were “blueish” in color, and they all had the same results — about 2ms maximum latency with averages under 1ms.
The storage array performance statistics are matching up perfectly with Figure #2, and Figure #2 is what I have been using for many years to analyze storage latency. Figure #1 (the "Top 10" graph) is something I normally have not paid attention to. However it is useful to be able to see the "Top 10" at a glance, so I would certainly like to trust this graph... but at this point it seems to be off by a factor of 100. Can anyone comment?
Figure #1. Datastore level >> Monitor >> Performance >> Real-time >> “Top 10” VMs in the vSphere Web Client:
Figure #2. VM level >> Performance >> Advanced >> Datastore >> Real-time in the vSphere Windows Client AND the vSphere Web Client:
