Home
| Calendar
| Mail Lists
| List Archives
| Desktop SIG
| Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings Linux Cafe | Meeting Notes | Linux Links | Bling | About BLU |
dsr at tao.merseine.nu seems to recommend *against* swap on RAID1: > With two swap partitions, you have twice the amount of swap as > if you RAID-mirrored it, and Linux knows how to alternate > requests from each partition, so there's even a performance > advantage. But in RAID1, Linux knows how to read the first sector that comes up (of the mirrored pair), so you reduce rotational latency with RAID1 (at least on the reads, not on the writes). It wouldn't surprise me if the RAID1 configuration gives *lower* latency (better performance) than the striped non-RAID configuration you're talking about. > Odds are good that if you have a disaster that takes out a swap > partition, you want to be rebooting immediately anyway... Depends on your application. If the server is supporting an office or development site, you'll want to defer rebooting until after peak hours. If the server is running a 24/7 commerce application, you'll want to keep it running without stopping. Motherboard IDE controllers won't cut it in that case, because they don't support hot-swap properly, but you can get inexpensive external IDE or SCSI controllers that can do hot-swap. And Linux RAID1 easily resyncs without rebooting. So you can build an inexpensive 24/7 server that never needs rebooting if you choose the tools and hardware carefully. But yeah, for the typical home computer, you'll probably find yourself tinkering with the system and rebooting it soon after a drive failure. I have two Linux systems at home, both with RAID1 partitions. I'll confess that on one system I have swap on the RAID partition, and on the other it's on a regular partition. It's nice to know that if I get a drive-failure alert on the first system, I don't have to rush into a reboot/recovery procedure--it can wait until after dinner or errands or whatever. And I'm thinking maybe I should reconfigure the second system. -rich
BLU is a member of BostonUserGroups | |
We also thank MIT for the use of their facilities. |