Hi
We are scaling up our HyperV S2D park. We already have two S2D HCI Clusters, each with 6 nodes. We now have 6 additional nodes ready to be brought into production.
So the question pops up. Add these 6 nodes to one of the two existing clusters ? Or, create a third cluster with 6 nodes ?
S2D supports up to 16 nodes, no issues there, although I find it more risky to have larger clusters. If there is ever an issue on the cluster level, you loose more workloads. Basically the 'do not put all your eggs in one basket' theory.
So what is the sweetspot for number of nodes ? Depends on your environment probably. It's more a gut feeling then solid judgement. Twelve nodes sounds too risky for me, or on the border of risk anyway. Don't know why, ask my gut.
So lets say I choose a 3rd cluster of 6 nodes. Then I get a 3rd S2D storage pool. And subsequently, this cluster will carry it's own RDMA/ROCE traffic for this pool on the same 10Gbit switches as the other 2 clusters. In short, I have 6 more nodes talking
RDMA in the same two storage VLANs as the other 2 nodes.
Is this a disadvantage ?
Should I separate the storage VLAN's for each cluster ?
I'm sure all options will work. On paper. So how do I get to making the right choice ?
Storage sizing isn't a factor. We don't use hybrid volumes with dual-parity, only 3 way mirroring. More nodes in one cluster will make the storage sizing more efficient if I was using dual-parity.
Any thoughts ?
Greetz
RW