We run about 65 VMs on a three-node cluster and are weighing alternatives as a hardware refresh approaches. The current setup uses shared storage provided by StarWind VSAN, with roughly 8 TB in use. We are considering Azure Local, Hyper-V, Nutanix AHV, Proxmox, and Scale Computing. What should we add to the shortlist, and what practical tradeoffs should we investigate?
What would you pick for a three-node cluster with 65 VMs?
Asked by
staticmarmot
Answers
For a small team, I’d compare the operational burden as closely as the license price. Proxmox can be a solid option, but make sure the people who will maintain it are comfortable with its ecosystem, backup workflow, updates, and support arrangements. Don’t judge it by a small test cluster alone. Exercise restores, node failure, upgrades, and the monitoring and alerting you would rely on during an outage.
Reply to this answer
Hyper-V is the most straightforward contender to investigate if you want to keep a conventional shared-storage design and familiar failover-cluster behavior. But check the exact storage and migration requirements rather than assuming it will behave like your current environment. Test host loss, storage-path loss, and a network interruption between sites if you have any stretched-cluster requirement.
Reply to this answer
Get detailed pricing and support terms for Azure Local before treating it as a simple hypervisor replacement. It may make sense if the surrounding Microsoft management and cloud integration fit your plans, but the hardware, subscription, and operational requirements need to be checked against your actual refresh budget. Ask the vendor to price your specific three-node configuration.
Reply to this answer
Nutanix AHV is worth a proper evaluation, but be clear about whether you’re willing to change the storage architecture or management model. There are configurations involving external SAN storage, though you should verify current support for your exact array and features. If keeping the existing storage approach is a firm requirement, make that a pass-or-fail item in the proof of concept.
Reply to this answer
Also consider whether replacing the platform is necessary right now. If the existing cluster is stable, compare the cost and risk of extending its life without support against the full migration and hardware-refresh cost. That is a business risk decision, not an automatic recommendation to stay put. For any replacement, include migration effort, recovery testing, and staff training in the numbers, not just host licenses.
Reply to this answer
Add your answer
Answers are reviewed before publication.
staticmarmot:
We’re not currently stretching across sites, so the main concern is keeping shared storage and HA without taking on a big redesign.