Who has migrated HCI from Xeon to Epyc?

OCSmash · January 21, 2020

I'm currently running a 2+1 Simplivity infrastructure (each node is dual xeon 16core (32 total) w/ 1TB RAM / 20TB usable storage) using VMware as my hypervisor. Essentially operating an HCI with 64 usable cores and 2TB of RAM in the production environment. I'm planning an upgrade to 96+ cores and 2-4TB ram. I've been mostly looking at the new Simplivity (HP) and VxRail (Dell) equipment. Both Dell and HP have introduced Epyc gen 2 options. This would essentially mean building 2x servers with 1x 64-core processor and 2TB of RAM each. I've even toyed with the idea of running one chassi with dual-Epyc 64-core processors and 4TB of RM and dumping HCI altogether. With the savings on VMware and other licensing, it seems like a slam dunk.

I have read that the migration process from intel to amd is not as simple as using vMotion to move everything from the old to the new cluster, which makes sense. We run mostly Windows Server and some Linux.

My questions is for anyone who has made this migration. How did it go? How did the OS images survive their migration form the xeon to the epyc platform? What was your performance like? Would you do it again?

Jarsky · January 22, 2020

I havent touched the new Epyc gear, but just keep in mind that nodes in a VMware cluster need to be the same CPU vendor type (i.e Intel or AMD). By switching between the 2, as you mentioned you arent going to be able to live vMotion the machines as Enhanced mode wont be available. To make the switch you will need to shut down each VM and move them over which will mean an outage. Otherwise I dont see any problem with the VM images themselves, i've switched between Intel Xeon & AMD Opteron just fine

As for running a single host, i'd keep in mind that while HP & Dell offer some excellent service, in the event of a hardware failure you'll be down for at least a few hours still until you get a technician out with the replacement part. Also keep in mind that you wont be able to perform regular recommended maintenance on the hosts such as firmware upgrades & esxi upgrades.

You might have more luck getting advice at Level1Tech as there is more people there that deal with SME's which would be more likely to make this switch. Most here are generally either enthusiasts (who typically dont deal with VMware) or Enterprise (who cant afford the downtime for a switch).

OCSmash · January 22, 2020

Thanks for your insights Jarsky. You make a solid point about the multi-host environment too.. We've had ZERO downtime with our current multi-host environment and there have been a number of updates performed over the years. HCI is a real life changer in the server room. I cross posted to spiceworks, but I'll check out Level1Tech as well. I've exhausted my personal & professional networks and no one had attempted the leap yet