Will it FreeNAS - CPU especially...

[overshoot]

Hi,

I am planning on switching my customer from FreeNAS (baremetal) to Proxmox->FreeNAS.
I can already hear some of you laughing out loud knowing how crazy I am but I am willing to do it.
They have a few Windows machine running as Servers for various applications and I would like them to become VMs of that main Server.

I have been running that same setup in my personal office for months as a test setup to find out if it was doable or not and it seems to be really stable.
I have a Dell R320 where they have a Dell T430 but we have 3 SMB users + a few VMs whereas they are 15 SMB users and will need a few VMs too.

Anyhow, here are the specs of the machine:

Dell T430
Dual Intel Xeon E5-2620 V3 (2x6 cores at 2.4Ghz, boosting at 3.2Ghz)
64GB of RAM
256GB NVME cash drive
Intel X540-AT2 (dual port)
MegaRAID SAS-3 3108 (HBA mode)
8 WD Red drives in a RaidZ2

They have had that setup running FreeNAS as an SMB Server only for a year now and it's been great.
They are using iMacs with a 1Gb/s NIC and they will switch soon to iMacPro with 10Gb/s NIC to speed up loading times and rendering also.

I am just curious to see what you guys have to say about the CPU since I've realized (too late) that SMB is a single core process that will benefit of higher clock speed CPU rather than a multi cores CPU.
Knowing that the load on the Server will spike once they will get the new Macs, I am wondering if the CPU won't become a bottleneck.

I have plan on providing them a quote with some extra RAM, an SLOG device and maybe a dual Intel E5-1650V3.

Please let me know your thoughts.
Thanks!

 

[anmnz]

You'd get a lot of agreement around here with the idea that it's worth getting as high a clock speed as you can for SMB, seems a good plan.

The "1" in "E5-1650v3" means it's for single-socket systems. You can't run two of them.

SLOG isn't normally useful for SMB unless you're planning to force sync writes.

More RAM is good. I'd go big on RAM and get very familiar with ARC statistics.

A significant client upgrade at the same time as moving to virtualization would scare me. Lots of exciting new bottlenecks to be discovered. Test, test, test.

Frankly this seems like a bad idea to me. Windows server consolidation sounds like a good plan that usually pays for itself, but do you really have to virtualize FreeNAS to achieve that? Can't you do the safe conservative thing and keep storage and virtualization platforms separate?

 

[overshoot]

First of all, thanks for your answer.
You have avoided me buying 2 CPUs...
But then the bad news is that I'll have only 6 cores (12 thread) for a virtualized Server having multiple VMs.
I have to look into that.

I would love to split the job with 2 Servers but it's about budget...
They have had that Dell T430 running as a Windows Server for a while and I saw the potential in it to have it running either ESX-i or Proxmox to consolidate the Servers + have their data stored on ZFS.
The first move was to replace Windows Server with FreeNAS and that worked well even though it's a Mac environment at 80%. That was the first challenge, especially since I have ran into issues when deleting heavy folders using the Finder. But it seems better since the latest SMB updates.

Having them buying an extra Server in these times knowing they are already upgrading the computers is going to be a challenge to have it accepted.

To be honest, I am feeling stressed out to do it as I feel the pressure but on the other end, it's been a working setup for us for a while now so why not?!
I get your point that it's risky to do all at the same time but the plan is to have the Server ready before they get the new Macs.
We should have 2 months to test it first. If it doesn't work, we can always go back to bare-metal since I'm going to share the mega-raid card to FreeNAS.

I thought the SLOG was useful for writes, when files are saved.
There would be no difference with a lot of RAM between opening files and saving them?