Storage: Stop Trying to Make Parity a Thing!

My order of 3 SSDs for my homeserver have arrived. My original plan was to get them into a ZFS RAIDZ, but I didn’t realize how much of a performance penalty there is. I thought the speed of SSDs would be fast enough for me to not care, but speeds plummet from the single-drive ~1200 MB/s read speeds to just ~ 400 MB/s!

I could definitely be doing something wrong - It wouldn’t be the first time - but I haven’t encountered anything special about creating a RAIDZ in my scouring of documentation and articles. It seems like the penalty for generating and writing the parity data is simply that severe/that CPU intensive.

I’ve given up on a RAIDZ for now. These SSDs were supposed to just act as run-time storage for my VMs, anyways; the VMs are still being backed up to other disks on a schedule. It would be nice to have some redundancy for my active VMs, but it’s just a homelab. Besides, 3 SSDs in a striped/RAID0 configuration is much more fun!

I’ll definitely be sticking with the tried-and-true Striped-Mirroed/RAID10 configurations from here on. More expensive in the short term, but the cost of one more hard drive is much less than the mental load of trying to get parity to be performant.