data storage

[da]

I appreciate reading about peoples' computer setups, even if they're not like mine. Occasionally I even comment. This time I'd like some advice.

I have a 700mhz PC running linux, which currently has a 200gb and a 320gb disk (*). I currently have about 160gb data. I'm probably not adding more than 50gb a year, at least as a guess.

I'd like to end up with a fileserver, since my laptop only has 100gb (**). I'm not mastering video or anything that needs top speeds; probably just serving music and video over ethernet or wifi b/g.

I want need some sort of built-in backup. rsync is an option, as is software raid. I like raid, it's super-easy. ...I could mount them both in the PC with software raid1 (limiting the raid size to 200gb). I would be happy with 200gb space for a while.

Or, I could buy another 320gb disk and mount the three disks raid5, with better redundency. Which is probably overkill.

Or, I could buy a non-pc-based enclosure, and turn off the PC, which would use less power. An openWRT router with USB hacked in would be ~$100 and sort of fun to put together. 5 watts instead of ~200. Then I could replace my ancient server with the 700mhz PC.

I don't like this answer because I've also been using the PC as linux development, stuff I don't want to expose on the public server. So that leans me back toward the current setup; slow server, faster dev box and file-server. (either off or idle a lot of the time).

Or, this might be the most elegent solution: put the PC in the closet and run virtual machines (vmware or xen). One vm would be a firewall/web-server/mail-host, with access to a smallish hardware partition. Another vm would be my file-server, using one of the above options, probably raid1.

My understanding is that IO, CPU, and disk are 95% as fast under modern virtualization. At lesat if I use dedicated disk partitions. This needs testing or verification by someone who's tried it.

With the vm answer, as a bonus, I don't need to mix the dev work on the file-server vm- I can clone and spawn new dev machines whenever I like. (er, which will chew lots of disk. Time for that terabyte?)

Thoughts?

(*) big thanks to bats22 who pointed out the absurdly cheap disk prices in a post last week. I had a disk failure yesterday (luckily, raid1) and I bought a new 320g disk for... $132 CAD including tax. Not too shabby, $0.41 a gb. Almost makes me happy the old drive died (though not really, since yesterday evening one of my tasks was swapping around disks and trying to find out why the server kept dying precipitously). I did my first-ever reiserfsck --rebuild-tree, which was only sort of scary. But it worked.

(**) Before this year did I ever expect I'd say "my laptop only has 100gb?" no I did not.