VMware Cloud Community
david_loup
Enthusiast
Enthusiast
Jump to solution

At which level is Datastore latency mesured ?

Hello,

I have a question concerning datastore latency. I see that, in general, 20 - 30 ms is the latency limit before having performance issue, but at which level is this limit ? is this the latency see from the VM ? see from Vsphere ?...

In Vcops, when I select a datastore and go to Operation - Details to see Read/Write Latency : Do I see Guest AVG or Device AVG ? (see image below)

Latences_Datastore.jpg

When I select a VM in Vcops to see the latency "from the VM", what is the difference between the vDisk latency and the datastore VM Latency ?

Latences_Datastore_2.JPG         Latences_Datastore_3.JPG

I know it's a lot of questions in one thread, so thank you in advance to people who will take time to answer.

Reply
0 Kudos
1 Solution

Accepted Solutions
mark_j
Virtuoso
Virtuoso
Jump to solution

Like the esxtop counters, you have different counters to measure from different parts of the stack. If you have latency at the disk level your queues will fill up and be reflected. Disk Command Latency is pretty good for picking up underlying storage latency, but it depends what resource you're looking at. Derived attributes of "latency" exist here and there that are calculated based on the type of storage/etc to make your life simpler and give you a single place to look.

When you start looking at queue latency, you need to keep in mind that is just focused on the queue. Go take a look at Duncan's article (ESXTOP - Yellow Bricks) on this and your questions will likely be absolved with regard to what BPs are and for what disk counters.

If you find this or any other answer useful please mark the answer as correct or helpful.

View solution in original post

Reply
0 Kudos
2 Replies
mark_j
Virtuoso
Virtuoso
Jump to solution

Like the esxtop counters, you have different counters to measure from different parts of the stack. If you have latency at the disk level your queues will fill up and be reflected. Disk Command Latency is pretty good for picking up underlying storage latency, but it depends what resource you're looking at. Derived attributes of "latency" exist here and there that are calculated based on the type of storage/etc to make your life simpler and give you a single place to look.

When you start looking at queue latency, you need to keep in mind that is just focused on the queue. Go take a look at Duncan's article (ESXTOP - Yellow Bricks) on this and your questions will likely be absolved with regard to what BPs are and for what disk counters.

If you find this or any other answer useful please mark the answer as correct or helpful.
Reply
0 Kudos
david_loup
Enthusiast
Enthusiast
Jump to solution

Thank you !

I understand better now.

Reply
0 Kudos