ZFS HDD Setup - What pool configuration for max speed?

It should be noted that ZFS has a bunch of cool shiny features you can use, like an SLOG and L2ARC. But in order for them to have any value at all, they must fit your workload. And for that you are going to have to sit down and figure out what that workload is in order to fit your hardware to it appropriately.

Slog is ONLY useful if you have sync writes. SMB and ISCSI by default don’t make sync writes. NFS generally does default to making sync writes. These can be changed. Other programs you use may or may not make use of sync writes. The purpose of an SLOG is to quickly and safely store a sync write so the program waiting for it to happen can continue. An SLOG will never be read from unless there is a system crash where the pending writes in ram are lost.

L2ARC, to vastly simply things, is only useful if your READ workload doesn’t fit in ram, but will fit inside the size of SSD you are using. The more complicated explanation involves careful (and error prone) analysis of arc hits and misses. Also note that ARC and L2ARC resist storing sequential reads, and try to focus on storing random reads. L2ARC is not to be confused with tiered storage that promotes and demotes data.

2 Likes

were any of them optane?

Oof I thought I toasted an optane that was only used for swap (and crash dumps) somewhat recently. Stopped showing up on the PCI bus entirely. Gave it a wiggle the other day and it came back to life though :thinking:

Back on topic though, congrats on figuring it out, whatever it was :slight_smile:

No. I was cheap.

1 Like

HI
How do you achive this result?
whats your setting of these disks

Thank you for this! First explanation that really helped me get an understanding of what’s happening.

Might know this page already and it may be out of date though still valid, quite interesting set of testing:

https://calomel.org/zfs_raid_speed_capacity.html

1 Like

I really dig that site, I should have suggested it here too

1 Like