0

I was reading this question where I felt obligated to comment and poison the sanity of system administrators everywhere. Then I had an interesting idea.

Lets say that you have a laptop with a one SATA and one NVMe interface. The only way to achieve a redundant array would be to pair these two very different interfaces (in terms of IO speed).

Instead of pairing identically-sized disks, what if you were to pair a 250GB SATA with a 512GB NVMe with a 250GB pool mirrored from the first disk onto a partition of the second. Then with the remaining space, set that as a read cache.

This would hypothetically provide additional IO for reads which would benefit from exclusive access on the faster device.

Zhro
  • 349
  • 2
  • 13

1 Answers1

2

Sure, it will work, though the pool will only perform as well as the slowest link in the chain (likely your SATA drive). Generally, however, you don't want to mix disks, much less a cache/storage disk. If you normally lose a cache disk, oh well, performance may suffer a bit, but you're no worse off. If you lose your NVMe disk in this scenario, you lose your redundancy as well. This is in a laptop though, so you're already limited WRT redundancy anyway.

If you're doing it as an experiment to prove you can, then I say go ahead and have fun, report your findings here and in superuser.

If you're planning on proposing it as an actual solution for anything outside of a lab at your place of employ...well...good luck having anyone take you seriously?

Also keep in mind that most SATA drives are almost always consumer drives. They usually don't support TLER, they have much less robust error handling (compared to SAS), and don't have higher quality components/firmware than their enterprise counterparts do.

Havegooda
  • 121