Hi, sorry if this is a duplicate topic. I am getting a new server for a bigger client and seeking advice.
I am planning to use Proxmox because I’ve used it before and like it, though I am open to discussion of alternatives (this goes for everything I say here!).
Database Server for Line-of-Business app (Server 2019, 16-32GB, 4-6 cores, 500GB disk)
App Server for LoB (Server 2019, 8-16GB, 4-6 cores, 750GB disk)
There are about 25 users here. The LoB app is a new one, so it’s hard to say how much a hog it is, but the vendor said the database server can get a little greedy.
I would like to get them a new Dell server, or a very good refurb. 1GB Ethernet is sufficient. I have been looking at the Dell T440.
Performance and reliability/redundancy is the most important thing (notice our storage needs are not very much). I would like this thing to FLY for the LoB app especially.
I have used Proxmox with ZFS storage in the past and liked it. I would plan to get lots of extra RAM for ZFS.
My questions are:
What’s a good model server to look at?
What would be the ideal ZFS configuration? HDDs or SSDs? What models?
Budget is $5-7K for hardware. I could push to $10K if I really had to. Given that, what should I prioritize? RAM? CPU? Fast storage? New server or refurb?
Any advice anyone can offer is much appreciated. I have decent experience but this is a bigger job for my scale. Thanks for looking!
you mention redundancy and reliability as most important but as I read it you are looking for a single server. I would suggest that you look at 2 smaller servers with something like a RasPI third node to give Proxmox the proper configuration for HA. I would go with SSDs over HDDs for your VM disks and Proxmox disks and have a look at ceph storage on the main nodes.
As for what hardware I would go with slightly used server hardware either by Dell or Supermicro which will give you the most power for your money IMHO.
As for CPU/RAM, you need to balance the 2 together to be most effective, as putting a super-fast CPU in with little ram will not perform well and same with the opposite. That being said as you are looking at ZFS and virtualizing I would bias slightly towards more ram.
SSDs are great for speed and reliability and I would look at using an array of HDDs as your backup storage for your VM Disks as backups can happen while the machine is running so they do not have to be super fast to minimize downtime when creating backups.
My business runs a number of Proxmox hosts and VMs and use older HP servers for the nodes with SSDs installed in the nodes for the Proxmox install and then all the VM disks are stored and accessed on dedicated storage servers and have had no issues with performance though we are running all Linux VMs with a few exceptions.
I run XCP-NG, but the concept is the same. Take a look at something like the Dell 730xd as you can get a pair of these for redundancy for a pretty reasonable price. Plenty of memory, CPU options and then fill it will SAS SSD drives and you will get great performance.
Full disclosure, they are a sponsor of my YouTube channel and you can use OfferCode LTSERVICES to get 10% off your order and yes, that is where we buy a lot of our servers from.
Tom and Astracea, thank you.
As far as redundancy goes, I think I would skip configuring HA on Proxmox. For this client, if one server fails, it would be acceptable to incur some downtime and restore the failed server’s VM(s) to the other host temporarily.
I’m looking at getting two R730s:
Xeon ES2630 8-core
64GB DDR4-2400 (4 x 16GB)
H330 controller (supports HBA mode I think?)
8 x 400GB SAS SSD; ZFS mirrored pair VDEVs
Let me know if I’ve gone horribly wrong anywhere. Thanks again!