I was about to pull the trigger on using VSAN for a VDI deployment (after spending an age choosing components), when I hit something I had initially overlooked.... configuration maximums say 100 VMs per host.
The design calls for 120 VMs per host, with an "emergency capacity" going up to 180 VMs per host. (Memory and CPU wise, the server can cope with this workload. This has been proven in a POC).
Is the 100 VM per host a "per host per disk group" limit? (i.e. can I split across two disk groups and get a maximum of 200 per host). Is there a very near future where this limit is increased?
The VDI deployment will be non persistent linked clones. VSAN licensing was included with Horizon, so would be a shame not to able to use it.
the problem is more in the amount of objects that can live in the vsan datastore. if all you use is a linked clone then each vm will use 3 to maybe 5 objects. any host adds 3000 objects to the datastore so going over the 100 vm limit shouldn't break things. once you start snapshotting vms and have multiple disks, replica's, FT's, and what-not, the amount of objects per vm rises quickly. i think vmware is being on the safe side when they say to limit the vm's to 100 per host. i hope i'm not wrong because we're stretching that limit somewhat too