mkfs performance.
Lots of people seem to be concerned that fsck is taking longer and longer as disks get larger and larger, but one thing that already is starting to be painful, that isn't mentioned so often, is creation of that filesystem in the first place.
Whilst mkfs should happen a lot less often than fsck's, mkfs of a 0.5TB ext3 filesystem on a 3 disk raid5 array already has me sitting around twiddling my thumbs. When we get to multi-terabyte disks, I might as well go shopping.
I wonder if something along the lines of chunkfs is perhaps the answer, where you mkfs chunks as you go along.
![[info]](http://l-stat.livejournal.com/img/userinfo.gif){kind=small}
kernelslacker) wrote,
