Slashdot Mirror
Best Redundant Storage for Home Use?
Brad Mace asks: "Despite my hard drive's dedicated service, I'm aware that it will someday fail. I'm not really interested in burning 100 CD-Rs to backup my hard drive, so I've been looking at RAID solutions. Obviously I don't need the best or the fastest stuff out there. What would be a reasonable setup for personal use? Have people had better experiences with internal raid arrays, external raid towers, or networked storage such as Snap servers? I'm primarily interested in low price and ease of use."
If you buy a RAID card with on board cache make sure the cache is battery backed up otherwise a power cut may corrupt your array. Some of the cheaper cards don't have battery backup.
Despite my hard drive's dedicated service, I'm aware that it will someday fail.
In terms of storage efficiency, nothing beats the naggy girlfriend:
The downside, though, is the insanely high maintenance fee. Noisy too.
raid only provides availability. if your fs gets hosed, you delete a file, or whatever you're unable to undo the damage. ditto if your pc gets burned or stolen.
if you want a real backup, make a real backup. if you want to do it cheaply, buy another drive, copy the contents of your data drive to it, and store it someplace safe. buying an external usb2/firewire enclosure will make this a lot easier.
I run a 6x120Gb software raid here. The redundency is nice but when you get that much storage in one place you start having other concerns.
Drive failures are scary. Note, if you're going with a hardware highpoint 1540 controller: forget raid 5. The array takes forever to rebuild (like 4 days) and sometimes fails midway though causing total data loss. (Other people may have different luck though, my friend is doing hardware on 4x160Gb and not enjoying it). I've had 2 drives fail on my software raid and the rebuild went well enough. I suspect the failures were due to inadequate power from the UPS causing a "brownout" condition when my power went off, since both failures happened after a power outage.
So other problems: thats 600Gb of disk space. Nobody in their right mind wants to sit through an fsck, so a journalling filesystem is needed.
but, all the journalling filesystems are new and untested. New your saying? They're a few years old.. and yes, thats true but they're new in comparision to all the other filesystems. And when you're talking about 600Mb of data on one filesystem it really starts to concern you.
(Even so, I've been running raid 5 with journalling on 4 to 6 drives for the last 1.5 years or so and haven't had filesystem corruption)
A friend of mine is running mixed drive sizes in his fileservr and doing no raid. He occasionally has failures and loses stuff, but at least he doesn't lose everything. Still, I'd be pissed if I lost an 80 gig chunk at a time.
The problem with raid is, you gain some redundency but you completely lose the ability to make sensable backups. Unless you're a corporation and can afford $3k for a tape drive (and an additional $500 for tapes) you're faced with the idea of mostly redundent, but no backups. Offsite backups become really appealing, but then you need to shell out the same money for drives that won't be put to use (until something happens)
I still don't have a solution for offsite backups.
As an aside, I started playing with encrypted partitions (not raided). That has the same sort of scaryness. One filesystem screwup and you lose everything. No backups unless you have a second hard drive. Then you're faced with raid 1 which would corrupt both drives, or copying the encrypted volume from drive to drive each night.
I've got no answer for that one ethier, except that encrypting a tape backup would probably be good in a corporate situation.
OK, my own experiences in this area:
I was looking for a basic, cheap RAID system to give me some redundancy.
I got an Adaptec 1200A RAID Controller card, and used two drives in RAID1 mode. This served me well, until I needed more capacity - RAID1 has 100% wasted space.
So I looked at RAID5. I got a Highpoint RocketRAID 454, because it was cheap. BIG mistake. The write performance, on my P4 2.8 Ghz, with three WD1200JB drives, was a terrible 9 Mb/sec, with 80-100% CPU usage. AVOID. I returned it then next day.
Now I have a Promise SuperTRAK SX6000. It's very nice, 25 Mb/s write with only 20% CPU thanks to RISC processor, but expensive.
In summary:
If you want RAID1 only, a nice cheap Adaptec will do fine. If you want RAID5, you will need a reasonable card. Promise do some cheaper ones than the SX6000, with less channels, you could look at those.
Hope this helps!
Right now, hard drives are the cheapest data storage medium bar none. Even an external USB or Firewire drive is cheaper per gig than tape/DVD/whatever for mortals once you factor in the cost of the drive (datacenters with terabytes of data are a different story).
For most of us, we have a lot of data that we keep "just in case", and that collection grows pretty slowly. So, once you've made the initial backup, the incrementals are pretty small.
There are three types of failure to worry about: drive failure, stupidity ("oops. I just deleted everything."), and major physical loss (theft, fire, etc.). Any second set of data on different media handles the first problem. An asynchronous second set of data (i.e. not RAID) handles the second problem. Off-site backups handle the third.
So, for off-site backups, I made an initial backup locally with a Firewire drive, then I shipped it to a friend's house. He leaves it plugged in to his machine, mounted off of my home directory on his machine. I make nightly incrementals and send them off to his place. Every few weeks, I grab the drive, take it back to my place, and do a fresh full backup.
If you have no friends (this is slashdot, after all), you could probably take the drive to work, so long as your boss doesn't object.
Forward, retransmit, or republish anything I say here. Just don't misquote me.