Hi,
So, I'm kind of new to the entire VSAN concept, and I'm looking for guidelines on a design for VSAN.
The environment in question is used for internal IT and is about to be converted to a Metro Cluster setup.
At the moment, we have 3 ESX hosts (HPE D380 Gen9) and 4 HPE P4500 nodes, plus a bunch of standalone hosts (Dell R720) with SSD-drives for those machines that need performance.
This is all good for the past requirements, but now we are in need to spread data across multiple sites and increase availability for internal servers.
Spreading the current P4500 nodes across two sites is quite easy, just spread them out and add a FOM on a third site and Network RAID10 will take care of the data, but what to do with the SSD-hosts?
My idea here is to add a fourth DL380 Gen9, add controller cards (HPE H240ar) to all four servers and add a bunch of SSD's (3x800GB) to them (plus add witness ESXi on the same site as the FOM resides on); this would give us replicated regular storage and SSD performance with redundancy as well.
I did some reading on VSAN sizing, and found out that in a all flash solution one of the SSD's would be used for caching in each host, and perhaps it's a bit overkill to have 800GB as cache and 1.6TB as data. I was thinking about adding a fourth SSD (480GB) to each host and have that as a cache drive.
Networking will be done with a fiber ring with 2x10Gbit on each link, so each DC will have 20Gbit between them in Mesh setup, so bandwidth is not an issue for this (latency is ~1ms, so we're in the clear for that as well). SSDs are for Mixed Use, so they should be fine. All of the HW are on the HCL.
Any inputs on the above solution? Am I thinking right about the sizing (I have no VMs to reference, we just want an estimate of the space we can use for VMs), or have I missed anything crucial?
/Andreas
