Linux's RAID6 Check Tool Sees 3x Performance Via Batching

For those that use Linux's raid6check command for checking/repairing an MD RAID6 software RAID array's device for errors, newly-merged code is showing up to a 3x performance improvement. The raid6check tool leverages the double parity of a RAID6 array to be able to error check and repair a RAID6 Linux software RAID device. Currently raid6check works on one stripe per disk at a time, while developer David Guillen Fandos took to batching it to speed up the I/O. David explained in the now-merged patch to the mdadm Git repository: "raid6check works at one stripe (per disk) at a time, which makes it not I/O bound. By batching I/O operations we significantly improve its performance at some reasonable memory cost (picked 512MB as a reasonable number, since it won't improve much more with bigger values, and it's a reasonable amount of memory to expect from a RAID machine these days). This bumps performance by 3x on a RAID6 setup (6 disks * 4TB) that uses regular spinning disks, plenty of memory and a modern kernel. I can only assume similar speedups for similar setups." A very nice improvement for the RAID6 checking tool performance for Linux software RAID.