Status on all-SSD systems?

[nurunet]

I want to replace an older QNAP NAS with 2x 4TB 3.5" magnetic drives with an ASRock Deskmini A300 to reduce noise and power consumption and boost Docker performance.

 

The Deskmini can fit 2x 2.5" SATA and 2x NVMe for storage. I got a good deal on two 2.5" SSD (SSD230S) and was planning to throw in another 2TB in NVMe storage.

 

Now, I read that SSD storage is not recommended for UNRAID arrays. This kind of throws a spanner in the works for me. I know there are TrueNAS Scale (but it looks like containers with Kubernetes are rather "involved") and OMV (but it felt a bit flaky when I tried it, having GUI issues on a fresh install), however, UNRAID looked like the best-suited offering for a small home server with some redundancy and good virtualization support - plus a very active community.

 

Now I read a bit, and also saw some buzz about the upcoming 6.12 possibly working more nicely with SSD, but I did not quite understand it. So I was wondering - is there a fairly straightforward way in 6.12 (or now already) to use 2x 4TB + 1x 2TB on SSD for storage - ideally with redundancy? Normally, I had hoped it to work out as an array with 4TB parity, leaving 6TB for data...

 

If running UNRAID means that either wear will kill my SSDs very quickly or that they will become unbearably slow, I guess there is no point.

 

Sorry for the n00b question, but the more I read, the more confused I got.

[itimpi]

The reason that SSDs are not recommended in the main array are:

• Trim is not supported
• Write performance is not going to be very good if you also have HD in the array.

what you CAN do is use a thumb drive to satisfy the requirement of at least 1 drive on the array, and then use the other drives in pools.    These can be BTRFS with current Unraid and the 6.12 release adds ZFS as an option for pools.   These can be multi-drive pools with various RAID options available if you want them to be redundant.

[SpencerJ]

2 hours ago, itimpi said:

what you CAN do is use a thumb drive to satisfy the requirement of at least 1 drive on the array, and then use the other drives in pools.

They touch on this briefly here:

 

 

[nurunet]

Thank you for your replies. What made me unsure about such a solution is that "cache pool" sounds like it was not meant to permanently store data. I'll look into it again.

 

One more question. One of the most charming features of the UNRAID arrays is that they'd allow combining multiple drives of different capacities for a redundant array with much more than 50% total capacity. Is this possible for a BTRFS or ZFS pool as well? Allowing me to combine 2x 4TB and 1x 2TB for ~6TB of redundant data? I'd currently assume no, only 4TB redundant and 2TB without redundancy?

[JorgeB]

2 minutes ago, nurunet said:

Is this possible for a BTRFS or ZFS pool as well? Allowing me to combine 2x 4TB and 1x 2TB for ~6TB of redundant data?

Yes with btrfs, no with zfs, but note that raid5/6 with btrfs has some known issues. 

[nurunet]

Well, I guess avoiding "known issues" is worth more than squeezing a tad more capacity out of my drives. Thanks!

[itimpi]

2 hours ago, nurunet said:

What made me unsure about such a solution is that "cache pool" sounds like it was not meant to permanently store data. I'll look into it again.

With 6.12 the terminology will I think be changing to simply refer to a Pool with one of the options being to act as a temporary cache for a share.  Technically it has been that way for some time, but for legacy reasons we tend to refer to the pool as the 'cache drive' as a few Unraid releases ago that was the only use for it.  With the move to using Unraid more as a application server rather than a pure NAS the pools are typically used to enhance performance of apps or VMs.   Also at that time only a single pool was available whereas you can now have up to 30.