How silly of an idea is it to use a single cache drive in an otherwise x2 dup setup?

[roasted]

Hi all. I'm testing DrivePool for what I hope will be a permanent setup here soon-ish. I have a server that's limited to 8 drive bays. The OS drive I have mounted inside and I was intending on doing 2x dup and starting with 4 spinning data drives. Originally I was thinking I would do two SATA SSDs for cache. This would leave me with 2 more bays to expand.

On the flip side, I also happen to have a single 500GB NVME SSD. PCIE adapters to mount single NVME drives are pretty cheap, so it got me thinking... if I do a single NVME cache drive I'd also be freeing up two more additional bays, thereby suggesting all 8 bays could potentially be for archive data on the spinning disks. This isn't a critical thing, but was a thought that dawned on me.

With that said, how dumb of an idea is it to hinge all ingest data on a single NVME cache drive with everything in archive being x2 dup? If it matters, I was intending on doing immediate offload in the balancer settings.

PS - side question, due to cost and my usage, if I do end up going with two SATA SSDs they'd likely be 250GB. What happens if I try to drop 300GB on the file server at once? Would it tank? Or would it simply allow the transfer to go and eventually max out the SSD cache drives and then lower speed throughput as it automatically begins to copy to the spinning drives? Just wondering if the size-of-cache = a hard set limit on how much you can transfer in one go.

EDIT - the more I think about this the more it concerns me, so I'm leaning back towards using two SATA SSDs for cache to ensure the x2 dup setting even applies to cache drives as well (as per my understanding anyway). The only thing is with getting two SATA drives it bumps cost up a bit and I'm on the fence between 250GB and 500GB SSDs. The big deciding factor will likely be the question in the PS field above (what happens if I transfer 300GB with 250GB cache? Does DrivePool gracefully switch over to using spinning data drives once cache is full or do I face a hard stop inability to transfer more than my cache size at any given time?)

Edited September 30, 2021 by roasted

[Umfriend]

AFAIK, but others may confirm, using a 250GB cache drive does indeed not work when writing a 300GB file. I think it is something that Windows/NTFS/DrivePool doesn't know in advance what the file size will be.

I would recommend using 2 cache SSDs. However, there is a workaround to use one physical SSD as 2 SSDs for caching purposes. It would involve setting up 2 unduplicated Pools, each using 1/2 of the SSD (which you would have to partition into two volumes, one for each Pool) and then combining the two x1 Pools into one x2 Pool.

It has its advantages (e.g., backing up using VSS becomes possible) but you lose a little bit of flexibility/efficiency and run the risk of the SSD actually failing. I have read here though that there is at least one other who does exactly this.