Every time the end users complain about the performance of an application installed on a VM, it becomes difficult for the VMware administrators to troubleshoot the problem. Today we will look at; datastore latency (read & write), one of the key metrics for monitoring application performance. Datastore read latency is the average time taken for a read operation from the datastore. Similarly, the datastore write latency is the average time taken for a write operation to the datastore
So when the time taken for the read/write operation to the datastore is high, the applications will not be able to function as fast as it is designed to which affects their performance. Many of the performance issues in virtual machines and applications traces down to the datastore latency.
This makes the datastore latency one of the most important components to monitor, so it doesn't affect the application performance and the end user experience. OpManager monitors the datastores using the VMware native API. Once vCenter is added to OpManager, all the datastores that are associated with the hosts will be discovered.
From the hosts, you will be able to view the read latency and write latency, which will show the total latency to the datastore.
If the write latency is on the higher side, you can find which VM is causing it, and take corrective action. Also set threshold for the latency, so if it violates the threshold, OpManager notifies you about the problem.
According to Karthikeyan, the read and write datastore latency is currently at 5 ms for the devices he manages and their applications are running without any problems. Numerous articles written about this mention if your datastore latency is above 20 ms, it will affect the application performance. So monitor your datastore latency, and if it is above the 20 ms, there are a series of blog posts by Joseph Dieckhans in this which shows how to troubleshoot the datastore latency issues.