measuring_disk_latency_powercli

[jvm2016]

Hi Luc,

i was checking following script to calculate maxIOPS .

Get the maximum IOPS - LucD notes

could you tell me how to get following expresion

$metrics = "disk.numberwrite.summation","disk.numberread.summation"

can we have similar for disk latency?

Thanks in advance.

[LucD]

If you look at the Disk I/O Counters, you'll find the disk.deviceReadLatency.average and the disk.deviceWriteLatency.average counters.

But be aware that they require statistics level 2 (for aggregate) or 3 (for device) when you go in the Historical Intervals

[jvm2016]

i run the following lines to check just to get realtime stats .

$metrics ="disk.devicereadlatency.average","disk.devicewritelatency.average"

$start = (Get-Date)

$vms = Get-VM | where {$_.PowerState -eq "PoweredOn"}

$stats = Get-Stat -Realtime -Stat $metrics -Entity $vms -Start $start

but got follwoingerror:

----------------------------------------------------------------------------------------------------------------------------------------

also could you suggest if the following code makes any sense this is not related to latency thing.

$tab = @()

foreach ($vm in (get-vm|?{$_.powerstate -eq "poweredon"}))

{

$vm.numcpu|%

{

$tab[$_.numcpu] = $vm.name

}

}

$tab

[LucD · Accepted solution]

For VMs you will need to use different metrics (the above ones are for HostSystems).

But be aware that they are only present in Statistics Level 4 (for Historical Intervals).

$stat = 'virtualDisk.readLatencyUS.latest','virtualDisk.writeLatencyUS.latest'

$entity = Get-VM -Name MyVM

Get-Stat -Entity $entity -Stat $stat -Realtime -MaxSamples 1

No, that doesn't really make sense.

I would do it like this

Get-VM | where{$_.PowerState -eq 'PoweredOn'} |

Group-Object -Property NumCpu |

select @{N='NumCpu';E={$_.Name}},

    @{N='VM';E={$_.Group.Name -join ','}}

[jvm2016]

Thanks. appreciate your help.

so i run to check for one of the vms.

and got following for realtime and max sample 1 .i think these values are normal as per my understanding latency more than  25 to 30 ms is point of concern.

could you comment on this.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

also could you tell more about following code specifically in blue line .

$lunTab = @{}

foreach($ds in (Get-Datastore -VM $vms | where {$_.Type -eq "VMFS"}))

{

    $ds.ExtensionData.Info.Vmfs.Extent | %{

        $lunTab[$_.DiskName] = $ds.Name

    }

}

[LucD]

A latency of 25-30 ms would be somewhat acceptable for classical HDD, but nowadays, with extended cache and SSD, I would say values over 10 ms might be of concern.

But again these are guidelines, and should be viewed in relation to your specific environment.

The variable $lunTab is a hash table.
I used it in those lines to map the CanonicalName of a LUN with the name of the datastore defined on that LUN.

The key into the hash table is the CanonicalName ($_.DiskName).