SLOG + Sata SSD for small VM host?

[oguruma]

Looking to build a SCALE box for a small newspaper office. I might end up using Proxmox instead, though I guess the same principles apply.

Will host a handful of VMs (web-based stuff like a CRM, development webserver, Nextcloud, etc) and serve as the primary storage server.

10ish employees total, so many 3-5 simultaneous users. Lots of medium-sized file transfer (photos), a small number of larger transfer (video).

I was going to create a pool of 3.5" rust for the main storage, and then a small SSD pool for the VMs/containers.

I'm looking at good cost/performance/redundancy solution to build a pool for VM storage. I'm leaning towards going with 3x 1TB consumer-grade SATA SSD in a 3-way mirror. Would anything be gained by adding a small consumer-grade NVMe SLOG for such a small setup? I'd assume probably not, since even with SATA SSDs that should be plenty of IOPS for such a small setup.

 

[ChrisRJ]

Your assessment looks right to me. Although the precise model of the consumer SSD will make a huge difference. As an example, a Samsung 860 Pro (MLC, rather good) and an 870 QVO (QLC, not good for write-intensive workload) are vastly different in terms of write endurance.

 

[oguruma]

Your assessment looks right to me. Although the precise model of the consumer SSD will make a huge difference. As an example, a Samsung 860 Pro (MLC, rather good) and an 870 QVO (QLC, not good for write-intensive workload) are vastly different in terms of write endurance.

I'm leaning towards the cheaper SSDs. The Nextcloud file storage will be located on the rust, so the SSDs would be mainly for the databases and things. Staying within the TBW for the warranty period for, say, a 1TB QLC SSD should be more than doable.

 

[Etorix]

In accordance with

The path to success for block storage

It seems like I haven't written a sticky for awhile, but just in the last week I've had to cover this topic several times. ZFS does two different things very well. One is storage of large sequentially-written files, such as archives, logs, or data files, where the file does not have the middle...

www.truenas.com

I'd go for a simple 2-way mirror for VMs rather than raidz1, but MLC or TLC rather than QLC. QLC is unlikely to fare well with many small writes and rewrites.

 

[oguruma]

In accordance with

The path to success for block storage

It seems like I haven't written a sticky for awhile, but just in the last week I've had to cover this topic several times. ZFS does two different things very well. One is storage of large sequentially-written files, such as archives, logs, or data files, where the file does not have the middle...

www.truenas.com

I'd go for a simple 2-way mirror for VMs rather than raidz1, but MLC or TLC rather than QLC. QLC is unlikely to fare well with many small writes and rewrites.

Thanks for the input. What's a good, future-proof way to do this? I'm half-tempted to get a couple enteprise NVMes (Kioxia, maybe?), or maybe something like the Nytro SAS SSD.

 

[Etorix]

That depends on the desired performance. For "a handful of VMs" serving "3-5 simultaneous users", I suspect that SATA/SAS drives could do.