Slashdot Mirror
HOWTO: 0.5TB RAID on a Budget
Compu486 writes "Inventgeek.com has a new how-to article
titled 'The
Poor Mans Raid Array.' The article details how to make a modular .5 terabyte
Raid 5 array for under $250 (USD), and it all runs on the Mandriva flavor of Linux." Drive prices being what they are, this seems cooler than it is practical. Update: 06/25 23:31 GMT by T : If that's not enough storage, Yeechang Lee writes "Let me show off the 2.8TB Linux-powered RAID 5 array I built for home use a few months ago. I provide lots of details on how I did it, what I used, and the results. The Usenet thread has good followup posts from others, too."
Actually, for RAID 5 you'd need a minimum of 3 drives.
n dependent_disks
http://www.acnc.com/04_01_05.html
http://en.wikipedia.org/wiki/Redundant_array_of_i
You are really better off just using software raid as provided by the operating system than using the fake raid provided by those on board ide/sata raid controllers. Then if your mobo dies you don't have to find one with the same raid chipset, worry about proprietary drivers etc. You just get another mobo and everything works fine. I played around with the nvraid, the silicon image raid, and one other brand, and they all pretty much suck. The best part is that without a special driver it doesn't matter how you configure the devices in the raid bios, they show up to the OS as individual drives not as a raid drive.
This is a bit off-topic, but I want to share my most recent experience with linux-raid
A few months ago, I decided I'd put together a RAID5 system in a dedicated box, to be used as network storage. I put together a Duron 1.6 on an ECS (I know!) K7VTA3, 512mb RAM, a Promise IDE controller, and 4 200GB drives. I figured the kernel-based software raid would be fine for my purposes.
I installed linux to a normal partition, then set up the RAID array. Everything seemed fine. I set up samba/nfs shares and ftp. Files seemed to transfer just fine. But for some reason, if I transfered a large file over the network directly to the RAID, the md5sum would have changed, no matter how I transfered it. To make things even more strange, if I transferred to a non-RAID partition, then directly used mv or cp to place it on the RAID partition, it worked great. Strange.
I never quite figured it out what was wrong, and I scrapped the project, with the intention to try again with some more decent hardware. Any ideas as to what happened?
What's with the .5TB? Is it not more standard to call it 512 GB, which, at least in my opinion, sounds far more impressive than .5 TB?
This is not economical, cutting edge, cool, nor is it practical. Why?
1) The drives are used. If you want to impress us, do it with new components with warranties (even refurb). Used makes it impractical and unreliable, even moreso because you didn't use hot swap.
2) It is only 500GB. This can be achieved in a RAID5 configuration with 3 NEW UNDER WARRANTY 250GB drives.
3) Heat. This negates the whole "cool" (both figurative and literal) label.
4) Power. Old drives suck up alot of power. Putting alot of them in a single case is going to draw a major stupid amount of power. Fewer drives can achieve the same effect with a reduced power draw. Did you take a page out of the AMD and 3dfx design methodolgy when you thought up this project?
Feed the need: Digitaladdiction.net
Nice project but.. at $15 for a 50gb drive 250gb raw will cost you $75. Add in shipping and I bet you are at $100+. I can get a seagate Sata 250Gb from Newegg for 120. I would rather have three of those RAID 5'd for 500Gb useable that some big, loud, hot, power hungry, loud drive array.
Many home RAID setups that I have seen are more for show than they are as a speedup. Most of them are RAID 0 - i.e., they're for "performance", not redundancy (I assume the reason is that most people don't want half their space to dissapear with RAID 1, but don't want to have to use enough drives to make RAID 5 effective). Yet the primary performance limits on a home PC's disk performance (apart from operating-system issues like the filesystem and caching) are latency-based, not throughput based; RAID, if anything, will increase your latency.
If you want a high performance system, spend the money to get a small, top-of-the-line drive for your root partition (15k rpm SCSIs are nice, if you have a scsi card - you can get a 9 gig for 30$ including shipping, an 18 gig for 55$), and then put all of your space-consuming files (movies, music, etc) on your cheap bulk storage. Get enough spare ram to have good disk caching. And, of course, choose a good filesystem for small files - ReiserFS works well for me, but there are a lot of good options.
You only need major throughput if you're doing a lot of very long file reads that need to occur at top speed (i.e., not playing video or listening to music; more like what you need for running a large relational database, or being a fileserver on a crazy-fast network). To the "Raid 0" crowd: Does this really fit your disk's typical usage patterns?
What a crazy random happenstance!
- Buy 3 250GB EIDE or SATA HD's very cheaply.
- Plug them into your cheap linux PC (with at least a 400Watt powersupply). If EIDE then make sure each drive is on its own (master) channel. If your BIOS supports "hardware" RAID, disable it.
- Use a low-level drive diagnostic fitness test to burn the drives in so you can be sure they won't fail right away. A great tool for this is The Ultimate Boot CD, as well as the 'badblocks' linux util.
- Assuming your 3 new drives are drives sdb, sdc, and sdd, with your bootdrive on sda (or hda), you should now partition each of them (instead of raiding the entire disk). I recommend creating one primary partition which is slightly smaller than the fullsize of the harddisk, such that if you buy a replacement drive of another brand and it isn't the EXACT same size, you won't be SOL when adding it. Mark the partition type as "FD", which is the raid autodetect type.
- Verify that your kernel supports software RAID by checking that
/proc/mdstat exists, or by checking for the multidisk "md" module in the output of "lsmod | grep md" after attempting to "modprobe md" and "modprobe raid5". If not supported, then... figure that out yourself.
- Now the fun part (assuming mdadm's installed):
/dev/md0 --level=5 --raid-devices=3 /dev/sdb1 /dev/sdc1 /dev/sdd1 /proc/mdstat
- Put a filesystem on the md0 device with mke2fs
/dev/md0 (or mkreiserfs, or whatever)
- Add a line to your
/etc/fstab to automount your new raid array at /raid5 or wherever.
- Oh, and if your distro doesn't automatically detect your array on reboot, you need to fix that by putting this in your init scripts somewhere:
Now, wasn't that easy?mdadm --create
View the status of the raidset construction by cat'ing
mdadm --assemble --scan
Power to the Peaceful
1. Get a big server tower case w/5+ 5.25" bays.
.7TB to boot. Or get a PCI EIDE raid card compatible with both Linux and Windows and go to town with RAID-0 and 1TB.
2. Get 4 250GB EIDE drives (cheap anymore!)
3. Get 4 $20.00 CompUSA lockable EIDE drive trays.
4. Get an SMP board + CPUs and slap 'em in there.
Ta-da. One power supply, four quiet drives, one case, software RAID-5 easily swappable with 2 dedicated fans per drive, looks professional, comparatively quiet, with the benefit of included scalable SMP workstation. And
There was a time when a SCSI array of many, many drives in a separate case at 10k RPM was something to lust after at home, but these days it just isn't. You can get close enough at home while saving space, using less power, and getting better overall performance.
STOP . AMERICA . NOW
That's exactly my experience. My family owns a PC shop, and we stopped buying WD drives around the time of their 30-60GBs because we were getting way too many failures - many DOAs and, worse still, drives that were failing a few weeks after being sold.
:-)
We've been selling pretty much exclusively Samsung and Seagate since then. The other *huge* often unmentioned advantage is that they're both much quieter than WD and Maxtor equivalents - Samsung being a little quieter in my experience.
Really, outside of the Raptor line, I see no compelling reason to buy a WD drive. I can definitely agree with the sentiment that I have *never* heard anyone say they hated Seagate drives, especially if you talk to the SCSI freaks out there.
Wrong.
No manufacturer is giving less than 3 years warranty.
The 10,000 rpm WD Raptor and all Seagate drives come with 5 year warranty.
I think you wanted to refer to the not so recent attempt by some major players to cut warranty to 1 year. That didn't last long, I guess because their sales must have suffered.
I'm surprised that nobody has mentioned the Linux Logical Volume Manager subsystem. It has many of the features of RAID arrays [such as spanning across multiple drives] with the added flexibility of being able to dynamically add [and theoretically remove] drives.
Unfortunately, aside from RAID'ing the volumes or something similar, I haven't been able to find any information on making the system redundant.
Read about it more on TLDP. It's a very robust system that works well on both servers and desktops.
Try: Comparison of nine Serial ATA RAID 5 adapters
Really, if you need that much storage, I would hope your hardware budget is a little bigger than what I allocate for my stuff at home.
Edward Burr
Having a smoking section in a restaurant is like having a peeing section in a swimming pool.
I think it would be more interesting to consider a redundent array of independent flash cards. Since it is clear that solid state drives will soon be included in PCs and laptops in the near future it would be nice to address the speed and reliability issues associated with them. This would also help with the heat and all.
Just a thought.
All your attention are belong to my old internet meme.
It depends on what sort of bandwidth you're talking about. Audio isn't nearly so throughput intensive as video. Plus, you're talking about writes, not reads.
Raid 0: Not quite double the read and write throughput; somewhat higher latency, but not usually a relevant amount. No redundancy.
Raid 1: Not quite double the throughput; slightly lower latency, but usually not a relevant amount. Redundant.
Raid 5: Very fast read throughput; lowered write throughput; higher latency. Redundant.
What a crazy random happenstance!