There's no one answer as to what is "normal" - it will depend on every piece of software and hardware between your apps/services inside the guest OS of your VM to whatever physical disk or disks are actually being written to.
As Scott said, there is not such thing as normal.
However, you can take a look to other VMs in the same datastore and compute cluster storage latency and see if the peaks you are seeing are replicated across all VMs.
You may wanna take a look at VMware Tools, virtual disk controller, virtual hardware version.
From the host perspective, you may wanna check connectivity issues, drivers or firmware versions and if the VM has the same behaviour in another datastore and another host. Also, since you are looking at Virtual Disk latency, the OS may be envolved, so I suggest you take a look at that also.
Virtual disk Latency is not the same as datastore latency. I think this document will be helpful for that Datastore Metrics
Also it's important that you are aware of backup processes or another I/O intensive tasks and when these happen.
Thank you very much for the answers. I want to ask one more thing. The datastore vrealize screenshot is attached. Do you think there is a data storage proplem? About I / O.
40 vm's are working on a single datastorage. we also replicate these servers to another site. Is it a normal value of 19.35 ms, does it have lower limit or upper limit?
If you want general thresholds to go by look here, this doesn't have exact write io latency but it has metrics you can get from the command line
and gavg would be the closest since its wat the guest sees as latency. 20 isn't terrible, but it depends on the application which was mentioned before, an application that processes financial transactions may be less tolerant of latency than a file server would be.