We're planning a refresh for a five-node VxRail cluster running about 55 VMs, mostly Windows. The current cluster has 53.5 TiB of storage.
A vendor proposed four vSAN Ready Nodes with 256 GB RAM and 256 GB NVMe for memory tiering per host, or five Hyper-V hosts with 512 GB RAM each and a traditional SAN. Is that memory difference realistic? What should we compare for licensing, resilience, and room to grow?
2w~bitcradle_948:
I'd also ask whether the four-node design still meets its performance target during a host failure, not just whether the VMs can start.