External RAID 5
September 7, 2007 7:29 PM   Subscribe

The amount of data I now have makes it impractical to continue backing up to DVD. I was planning on building a RAID 5 array, but I'm in over my head.

The computer doing the hosting would be an older PowerMac G5, which means PCI-X. It would be sharing over my network (gigabit). I have over a terabyte of data to store, so I was planning on buying 4 500gb drives. In case you've never looked inside a PowerMac G5, there are only two drive bays, so the drives would have to be external.

So how the hell do I do this? I've already seen the Drobo and I'm not too interested – it's $500 and butt slow. I'm totally willing to spend $500, but not on something that slow. I don't need the array to be super high performance, but 11mb/sec isn't going to cut it. Also, their marketing gets on my nerves ("data robot"? come on.)

I've looked into some cards, but I'm not sure what to get. External SATAII connectors and an enclosure? Most of the manufacturers I've seen so far (areca, high point, etc) have really sketchy websites and vague compatibility information.

To sum up:
Need RAID 5 solution, drives must be external.
Must be compatible with PCI-X, Mac OS X.
Everything would be plugged into a UPS.
posted by tumult to Computers & Internet (15 answers total) 7 users marked this as a favorite
 
The NORCO DS1220 (newegg) is sweet. I just built one today, 12 500 gb Samsung SATA dives, Mac, Windows, and Linux support. Has 4 eSATA channels for 3 gb/Sec transfers, OS RAID, JBOD, or Hardware RAID. Total cost for the chassis and 13 SATA drives = $2384. 6 Tb + 1 spare drive.
posted by nikko at 7:44 PM on September 7, 2007


I just went down the raid-5 route - it was ok, but more trouble than it was worth with either FreeNas or Ubuntu. I ended up getting a PC with 4 drive bays at our work pc sale, installing 2 500 GB drives (space for 2 more when I need them) and doing rsync cron'd to back up the data. 4-5 hours of risk is something I'm willing to take.
posted by true at 8:06 PM on September 7, 2007


DERAIL

Make sure you're clear about your desired result. Redundant disks are good for uptime (continued data availbility in the event of a disk failure) but bad for backup in that the tend not to be very portable. If all you're concerned about is disk failure, go for it. If you need to get it off site ("site failure," let's call it) and care about archival quality (readable in 3 years or more) you should probably be moving away from optical media anyway.

/DERAIL
posted by ZakDaddy at 8:52 PM on September 7, 2007


ZakDaddy speaks the truth RAID5 doesn't protect you from corruptions, accidental or malicious deletes or some catastrophe (fire, flood, theft). A big RAID5 NAS solution is good and you may well need it (I am watching thing thread because I need something like it too) but this doesn't really absolve you of the need to have a backup plan in place.
posted by mmascolino at 8:59 PM on September 7, 2007


It seems like RAID 5 is overkill, if this external scenario is just the backup. If it's the primary storage, then it might be justified.

How about a FW 800 RAID 0+1? Plug and play.
posted by Devils Rancher at 9:02 PM on September 7, 2007


First, let me echo what everyone else is saying: RAID isn't a backup. RAID is meant to prevent downtime, not data loss. It's quite easy to have two drives fail out of one RAID... much more than you'd think, because if one drive goes early, there's a much higher chance that the others will have the same manufacturing defect. I've had it happen a couple of times, and it isn't fun.

What you appear to need is both a ton more space AND a backup solution. The space part would probably be best solved by a dedicated NAS box -- a dedicated server, basically. (NAS means "network-attached storage": a network file server, basically.) You're not highly techy, you need a crapload of space, and you need it to run as reliably as possible, so not rolling it yourself is probably a good idea. I've heard good things about the Infrant ReadyNAS boxes -- the base units are abour $1K, though, which is more than you wanted to spend. (That Norco unit that nikko recommends seems to be getting pretty bad reviews on Newegg; how well that reflects the actual product quality, I don't know.)

From there, I would suggest adding individual drives to your Mac and using them as backup. Usually, when you get a big array, there's a bunch of crap that you don't truly need to copy, and then a core set of files that would kill you to lose. As long as your core data will fit on one hard drive, just do a network copy fairly frequently down to your Mac; if your core data is small enough, you can do several generations. And if you want offsite backup, buy an extra drive and rotate. Don't forget to also back up your workstation(s) to the RAID.

If your core storage needs get bigger than the biggest single drives you can find (currently 1 terabyte), then you'll have to either split onto two drives and spent more attention backing up, or else go to a tape library. They are obscenely expensive.

The big advantage to a NAS is that it's cross-platform, so you can use it with any kind of computer you want without having to worry about compatibility. If you want to add a PC or a Linux box to the network, it's trivial. If your main machine crashes, no biggie, you've got everything on the RAID and can rebuild quickly. If the server crashes, no biggie, you've got your most recent backup and can rebuild it.

But when your server and your workstation are the same box, a crash can really, really mess you up.
posted by Malor at 9:22 PM on September 7, 2007 [1 favorite]


Here at Fort Awesome we have a Terabyte of data kicking around that is in a mirrored disk arrangement, but really only about 40 Gigs that we MUST HAVE in the event of something bad. We offsite over the network this data and have it back up revisions and incremental changes to the MUST HAVE data that amounts to 3-4 MByte a day.

The key with data storage is to seperate the MUST HAVE from the gee, what a pain to get it back to normal. For the Fort Awesome household we're comfortable with losing the majority of our data because it's mirrored and unlikely to go walking without notice, to the extent that in our mirrored set up we keep an onsite spare of each drive and model in an array in case one drive goes tits up so we can rebuild the array quickly.

The net of it is it's about 400 bucks for spare unused disks still in the box that we keep around in case of failed array member, and we pay 5 bucks a month for each machine to back up to Mozy.

Really what you're talking about would be commercially a DASD, high performance raided storage, those things are going to be out of your range. If you can survive it maybe consider a striped and mirrored set up and offsite or use a Terastation for the incremental backups.
posted by iamabot at 9:34 PM on September 7, 2007


Just chiming in to note that your media storage doesn't need to be as fast as your primary computer startup disk. Even a disk or array of modest speed is probably okay to stream media from.
posted by scarabic at 9:58 PM on September 7, 2007


Response by poster: Malor: Actually, I am highly techy :( I just don't know much about RAID solutions. I already do critical backups of data and store them elsewhere. I need an assload of space to store all of my projects while I'm not working on them, but which I might change in the future. Like, I have a 50gb audio project I'm working on, I need to be able to move it off of one of the internal drives on my workstation because one of the internal drives on it is filling up. But I might need to go back to that project within the next couple of weeks, so I don't want to have to burn/tape dump the whole thing off multiple times. I can just copy it to/from the networked array (working on the file remotely would be way too slow.)

Also, my workstation and server are not the same computer. The PowerMac G5 is ONLY a server (DHCP, firewall, bittorrent client, ssh etc.) I have 6 computers on my network.

Anyway, all of the internal bays on all of my computers are full. Audio projects at 24/96khz get really large really fast.

ReadyNAS is cool, but I really want something attached directly to my server. I mean, it's already doing SMB/AFP shares of other drives attached to it, and I have it administered the way I want already (write access only from encrypted connections from itself and my workstation). Plus, yeah, ReadyNAS really expensive.

The Norco thing seems kind of iffy, I guess. I wouldn't need a 12 bay array, and the 4 bay one seems like it's missing some important features. But thanks everyone for the help so far.
posted by tumult at 11:25 PM on September 7, 2007


doesn't help the original poster (sorry), but related to comments above. i have software raid5 running in suse linux and it was easy [self link] to set up - the hardest part (once i understood what i was doing) was finding the right menu in the installer interface. i have tried measuring speed and, as far as i can tell, it's not "really slow" on writes, just no faster than an ordinary disk (i am no expert at this kind of thing, but i ran some benchmarks and could see no difference).

i wonder if building a cheap linux box might be your best approach? it's easy to find mobos with 4 sata ports. this machine was a day's work (iirc) and i am really happy with it.
posted by andrew cooke at 1:36 AM on September 8, 2007


Response by poster: I'm actually seriously considering the ReadyNAS NV+ now, despite being it horribly expensive. I just can't find a satisfactory, consumer-rish priced solution that met my original criteria.

Thoughts? Also, has the acquisition by Netgear impacted the quality?
posted by tumult at 3:15 AM on September 8, 2007


Response by poster: So I originally thought the ReadyNAS NV+ was about $650. Apparently, I was wrong, and Netgear has raised the price to $850 since acquiring Infran. So, the ReadyNAS NV+ is out.
posted by tumult at 5:21 AM on September 8, 2007


Areca's website might be "sketchy", but their hardware and software is among the best, from what I hear (mostly from FreeBSD people). They charge a premium for it too, so maybe a different make would be more appropriate for a budget.

Anyway, an external SATA/SAS box like this and a good external HBA like this (recent PowerMac G5 drivers) should do the trick, though the Areca cards seem to start at around $379.99, which doesn't leave much for the enclosure. My next port of call would probably be LSI and a Fusion-MPT based card, which are very mature and supported pretty much everywhere.
posted by Freaky at 6:45 AM on September 8, 2007


Ok, so if you are techy, that changes the equation quite a bit.

Fundamentally, your problem is that the Mac just doesn't have enough space or ports for enough drives. Further, even if you figure out how to ADD enough drives, you can't do software RAID5, so you'll have to buy a hardware controller. They cost a lot of money, and then getting one with drivers for the Mac can be tricky.

My suggestion? Scrap the Mac. It's a fine machine, but it won't do what you need it to. Further, even if you add the stuff you want, it may not have enough internal bandwidth to keep up with Gigabit Ethernet, which you are going to want to look into.

What you really want is a PCIe-based machine running Linux, using software RAID5, for maximum performance at minimum cost. Linux software RAID is quite good, especially on a dual-core or better system. You should easily be able to build a server for $750 or less, including a good case with lots of drive bays. Make sure your motherboard has as many SATA ports as possible. Buy a small and cheap drive for the system boot drive; you could potentially add a second for a RAID-1 mirror if your chosen motherboard has enough ports. Then add however many drives you want for storage. Finally, buy a single monster drive for backup.

This will, obviously, take significant time to set up, but it should be able to stand in for your G5 in every way. Pull all the services except firewalling over the Linux box; keep the G5 to do your firewalling, or replace it with a separate unit. It's never, never a good idea to have your firewall and your fileserver be on the same machine. (Ideally, a firewall should be locked down to a ridiculous degree and do ONLY firewalling PERIOD.)

Once you have your Linux box up, get a Samba share going (Windows-style networking). if you have other machines in the net that have gigabit Ethernet, look into one of the cheapy Dell gigabit switches, which aren't terribly expensive and perform quite well.

With a Linux box running Samba on a software RAID that's not choked by PCI, over Gigabit Ethernet, running directly from the server should feel pretty much like running from a local drive. You shouldn't have to do much, if any, transferring from machine to machine just to work on a file... and even if you do, the GigE should speed it up a bunch, if your machines are fast enough to take full advantage.

The downside to this solution, of course, is that you'll have to put a lot of effort into building and learning the system, but if you want maximum value for your dollar, this would be the way to go.
posted by Malor at 8:06 AM on September 8, 2007


I run a RAID 5 file server at home. I run many RAID 5s (and RAID 6s) in the office. I also admin older Powermac G5 systems.

I wouldn't spend any amount of time doing RAID 5 with a Powermac G5. The early PCI-X ones are weirdly temperamental with drives and RAID controllers (early models aren't compatible with Western Digital Raptor SATA drives out of the box, for example, which was certainly a surprise). If I were in your shoes, I would buy and/or build an external SMB/NFS server.

However, in this scenario, there's a very simple "cheesy" solution that hasn't been brought up. Buy two known-compatible dual-drive RAID-1 enclosures (from OWC, who take special care as to Mac compatibility with every drive enclosure in their store), and outfit them with 4 750GB drives of your choice. (You can get them pre-assembled from OWC, but often at a premium unless they're on sale).

That's 1.5TB of storage in two 750GB RAID 1s. Your G5 will recognize them like any other USB or Firewire drives, except each drive will silently be mirrored in their fat little enclosures. Or, get four 1TB drives, and have two 1TB redundant enclosures.

Just a thought.
posted by I EAT TAPAS at 10:43 AM on September 8, 2007


« Older Weird shortwave radio signal   |   Brand new noise coming from G5 into audio signal Newer »
This thread is closed to new comments.