Do you mean you have 3 physical nodes, each of which has 16Tb.
> How much storage is considered good for every single vm
How much do you need? I don’t understand the question. Only you can figure out your requirements.
> separate volumes for os installation and hdfs data
Well, yes, that is wise, but remember that MapReduce spill goes into the “local” disk space, not HDFS, and so that is typically /var on the OS disk. You don’t *want* spill if you can avoid it, but sometimes you can’t.
Also consider separate partitions for logs different from the OS ones.
> 30vms(slaves-workers).
That seems like too many. How many CPUs do your nodes have? How much memory?
Goodluck
