Write throughput vs latency on Hybrid

[andvm]

Hi,

On a typical Hybrid setup assuming all components follow the VCG, would one expect a VM to incur a Write latency of around 250ms when writing at around 70MB/s ?

I can see this from the VM / Monitor / vSan / Performance by looking at the Latency Graph.

FYI - Skyline Health shows all Green and Healthy, think this Write to Latency "issue" has always been there but never understood if this is within expected limits or not as do not think it is straightforward to draw a line or graph to show average expected latency at a certain throughput (MB/s) as it probably depends from many factors.

[Deso1ator]

No, that latency is definitely abnormal. Is it a quick spike or sustained?

[depping]

Is this during performance tests or normal operations? 250ms is never normal, unless you are pushing the system like crazy, and you have an extremely small write buffer etc.

[andvm]

normal operations, whilst the backup is running on the VM.

Any further checks (graphs/commands) I can do to find out what is the bottleneck causing this. Could this relate to large block sizes used by backup application therefore causing high latency? 

We do not have vROPS as that might have helped.