Raid 5 (was: Mounting of Linux volumes)
torriem at chem.byu.edu
Thu Dec 1 18:45:28 MST 2005
On Thu, 2005-12-01 at 16:56 -0700, Ross Werner wrote:
> Since we're on the subject of RAID, what's the speed difference like
> between RAID 1 and RAID 5? Is that pretty much the only benefit of RAID 5?
> RAID 5 seems nearly impossible to recover from anything apart from a
> single drive failure--any sort of data corruption, or a power outage, or
> pretty much anything unexpected can render your data unsalvageable,
> especially if you're using software RAID ... whereas with RAID 1, you
> always have the option of just treating the disk like a normal drive.
> Are there compelling reasons to use RAID 5 over RAID 1? The convenience
> factor of RAID 1 just seems to outweigh all performance benefits of
> RAID 5, especially when you factor in recovery time and things like that.
I use a combination of both RAID 5 and RAID 1, and I do full backups
nightly (well of certain things), in the event of catastrophic failure.
My main file server has two RAID 5 arrays mirrored together in a RAID 1
configuration. I feel fairly comfortable with that arrangement.
I have also found in my work that sometimes a periodic syncing or
cloning of a disk is better than RAID 1. For example, my other servers
all have 2 disks in them. Once a week a script runs that rsyncs the two
disks together so that if one fails, I can simply boot it off the other
drive. Of course this means some down time. But the periodic rsyncing
has the advantage of giving you a known good image to roll back to if
something messes up. In RAID-1, if you get compromised or somehow
scramble the data, it is scrambled on both disks. So if you don't have
data that changes too often (in other words you can afford to lose a day
or so of data back to the last sync), and you can afford a small
downtime, then this might be better for you than RAID-1. I'm going to
be building a new file server soon and I think I will be going the
syncing route and sync the two RAID-5 arrays nightly. Another reason
for this is that although losing a disk is a concern to me, it's even
more likely a user will call up and say "I deleted a file, can you get
it back?" With RAID-1, I can't do it without going to the backup
system, which we, as a matter of policy, won't do unless we have to
rebuild the entire system. With the syncing, I can with minimal effort
get go back day or so (depends on the schedule). Even with a 7 TB disk,
I think a nightly rsync may only take a couple of hours, since the data
change set isn't that great each day.
On a related subject, what do each of you do for backup? As I get ready
to build the 7 TB machine (brand to be determined), backup is a huge
concern. Tape is too expensive. Removable disks seem like my only
Another issue is archiving. Do any of you have an archive system (as
opposed to backup) where I can go track files all the way back to their
creation? No a version control system will not work (although a
versioning file system like veritas might). I am wondering how hard it
would be to crank out a DVD changeset every night (a few dvds a day is
reasonable), combined with a database to track which disk has what file.
> ~ Ross
> PLUG: http://plug.org, #utah on irc.freenode.net
> Unsubscribe: http://plug.org/mailman/options/plug
> Don't fear the penguin.
Michael Torrie <torriem at chem.byu.edu>
More information about the PLUG