On Friday 14 November 2008, Henrique de Moraes Holschuh <hmh@debian.org> wrote about 'Re: What is the point of RAID?': >On Fri, 14 Nov 2008, Boyd Stephen Smith Jr. wrote: >> On Thursday 13 November 2008, "Douglas A. Tutty" <dtutty@vianet.ca> >> wrote >> about 'Re: What is the point of RAID?': >> >The other thing to consider as that you don't necessarily need the >> > same performance/protection for the whole dataset on a system. >> >> Yeah, I can understand that. I use software raid-1 for /boot, software >> raid-0 for (a) directories I consider "owned" by package management, so >> that reinstalling from fast net or local mirror solves that problem and >> (b) "throw-away" data like /var/tmp, and hardware raid-5/6 (depending >> on # of drives) for real data like /home, /srv, /var, etc. > >At which point you might as well use two or more swap partitions with the >same priority (and _not_ on top of a raid device), to get high-speed swap > as if it were raid-0'ed. After all, a single drive crash WILL bring > your system down anyway... Yeah, I'm not doing swap that way, because I like being able to resize swap as easily as any of my other logical volumes, so it's on LVM on RAID-0. I find if you want swap on LVM, it's better to put RAID-0 under LVM rather than multiple logical volumes used as swap at the same priority, that that could be different with a different setup. (Yes, my way is a slower.) My current setup: 2x 74G 10k disks -> partitions -> md0 (raid-1), md1 (raid-0) 8x 1T 7200 disks -> Areca -> sda (raid-6) sda + md1 -> vg -> 9 lvs. Not sure the best way to convert that setup to using two lvs (one on each 10k drive) as same-priority swap. BTW, this is a desktop, there's no way I suggest this for a server that has to be up in the case of drive failure. For a server, raid-1 or raid-1/0 on fast drives [2] for the "system files" and raid-5 or raid-6 if there need to be a lot of attached data. "A lot" is more than 2 large HDs worth, less than that but more than fits on the system drives would be a second raid-1 or raid-1/0 on large, slow [1] drives. A single disk failure in a server should not stop services and should be recoverable without a reboot. [1] < 10k rpm [2] >= 10k rpm -- Boyd Stephen Smith Jr. ,= ,-_-. =. bss03@volumehost.net ((_/)o o(\_)) ICQ: 514984 YM/AIM: DaTwinkDaddy `-'(. .)`-' http://iguanasuicide.org/ \_/
Attachment:
signature.asc
Description: This is a digitally signed message part.