But AnyRaid-Z1 would eliminate that by giving you a total of 18 TB—the complete usable space for that set of drives. It will do so by splitting each disk into 64 GB chunks and then treating each one as a single standalone device. Think of it as splitting your drive into multiple 64 GB partitions, and then assigning them all to a conventional pool.
I did this as an experiment, using larger partitions of 1TB. It worked well as such, but the key issue is scrubs and rebuilds.
ZFS doesn't check which device the partitions are on, and treats them as independent. Thus during a scrub or rebuild it will issue at least one IO per partition. ZFS tries to issue sequential IOs for speed, but due to the above the drive sees a bunch of random IOs, and the scrub/rebuild speed drops like a rock.
Curious how they will address this, I presume with a layer inbetween.
But AnyRaid-Z1 would eliminate that by giving you a total of 18 TB—the complete usable space for that set of drives. It will do so by splitting each disk into 64 GB chunks and then treating each one as a single standalone device. Think of it as splitting your drive into multiple 64 GB partitions, and then assigning them all to a conventional pool.
I did this as an experiment, using larger partitions of 1TB. It worked well as such, but the key issue is scrubs and rebuilds.
ZFS doesn't check which device the partitions are on, and treats them as independent. Thus during a scrub or rebuild it will issue at least one IO per partition. ZFS tries to issue sequential IOs for speed, but due to the above the drive sees a bunch of random IOs, and the scrub/rebuild speed drops like a rock.
Curious how they will address this, I presume with a layer inbetween.
There's a rewrite command now, it defrags and does all kinds of nice things for your array.