How do you feel about rolling your own, for God, country and savings? Zonk mentioned the Proliant Microserver. It's a tiny, quiet little thing with 4 x SATA slots for your drives and some extra external ports for expansion. They also running regular offers for cashback on purchase. Current one in the UK is £100 back off the £160 purchase price of the machine. I'm betting there will be similar deals in the States. The box is stupid cheap either way. 4 x Seagate 3TB drives gives you 12TB raw for around $400 4 x Seagate 4TB drives = 16TB for around $640 So that's 12 or 16 raw TB for $500 OR $750 ? On my Microserver I installed FreeBSD to a USB stick and boot from that. Then I configured ZFS from the command line and created my pool. If you don't want to fiddle around with FreeBSD, there is FreeNAS which does all the goodness in the background and gives you a nice GUI for configuration. The benefits of a little multi-purpose box, though, is that you can have it do other things in addition to providing NAS. Mine grabs torrents, handles networked security cameras in the house, lets me get back into the LAN when I'm travelling etc. Also it's extremely low power, which saves a few pennies when you're running it 24/7 for the year. One thing I'll get around to doing eventually is trickling an entire mirror of the data to Amazon S3 storage for a proper offsite backup.
That's what I did last time. It was great until it stopped working abruptly and painfully. Then I was left to fuck around on my own. Don't get me wrong - I was inches away from a DL380 a few years back. I'm not opposed to the idea, I'm just not sure what I benefit from going that way here. But work it out with me, though. So the box in question is probably this one, yes? That's $339 raw, 4 drives. Whatever I want on it, I put on it and make it happy. For another $10, I get this, which has hot-swappable cradles, the possibility for redundant PS, and a design ethic revolving around "people who don't want to deal with configuring a NAS." It actually works out cheaper - because WD wants you to buy drives, a fully loaded 16TB EX4 comes in at a "raw chassis" cost of $280. Or, I bite the bullet and spend another $400. The equivalent of two Proliant microservers. But what I get with my hard-earned change is this - now we're 5 bays, not 4 (which, if you're going to do RAID5, is your multiplier of choice). I've got dual power supplies. I've got dual fans. I've got four aggregatable ports. I've got expansion chassis that will push this bitch out to 15 drives if I feel like it. I have the ability to cache on an SSD. I've got RAM I can upgrade. And I've got a web configuration that, while kinda scary, doesn't scare me like "FreeBSD" and "ZFS." Again, I built one of those in 2003. It lasted three years and then failed dramatically, leaving me shit out of luck. Due to life events, cash flow and cantankerousness it took me nine months to get my data back. Color me stung. I would appreciate your opinion on this. From my perspective, it's worth the $400 hit to have the expandability, to have the support of a non open-source organization behind me, to have a single-point vendor, to have an optimized configuration, to have the ability to build a 16TB array in one chassis rather than a 12TB, to not have to puzzle out all this stuff. My perspective can be changed, however. That's why I posted this. What's your opinion? By the bye, fun power fact: Our old apartment is 3 doors down. It's a 2 bedroom vs. a 3 bedroom, though. Nothing else has changed, other than the fact that the refrigerator in this apartment isn't fully enclosed in wood, which means it doesn't run all the time. My power bill went down $80 a month.
Well, true 'nuff, the argument about how much one's time is worth comes into play. At the time I set up The Box, I had spare time on my hands and I used it as an excuse to learn FreeBSD. If I had to do it again? Maybe I'd spend the money on (or install) something more fire and forget. A thought: you can get 5 drives in a Microserver by stripping out the CD bay. Again, time. Another thought on RAID5: from a discussion about RAID and large disks, I can quote my IT guru here:
"As always, this is a standard risk calculation. If the data matters, you don't use RAID5. And with big SATA drives you are often best off using RAID6 (even with 4 disks) as it has the highest recovery coverage." I can ask him to explain further why this is the case, but I'm assuming double error parity is better than single for recovery. As with everything, it comes down to you usage. For me, it's long term backup of large files (video data) that I don't need online access to, since the projects go onto faster local storage when in use and then move back to the slower, larger NAS when packed away. Do you need bandwidth critical storage? Or if it's primarily Time Machine backups, your first sync is the one that takes the time and then it's small, incremental updates. In that case, maybe it's slower, cheaper drives and a slower interface. Like you I have nearly no idea what I'm doing. However, i's educational doing it. ZFS though? Proper space technology. Raid-Z (1, 2 or 3 vs RAID5, RAID6 or "RAID7"), self-healing data, automated snapshot and rollback of data, on the fly storage compression. If I had to do it again (or I upgrade The Box in the future), I'd go with FreeNas 9 or Nexenta Community to avoid having to muck about in BSD. Open source, yes, but still reasonably turnkey, and they both appear to provide the latest ZFS implementations which older FreeBSD versions do not.