What is Digg?47 Comments
- Silby, on 10/12/2007, -9/+32isn't RAID for killing bugs?
- rguenthner, on 10/12/2007, -3/+16Redundant Array of Inexpensive Disks - or did this change when I wasn't looking?
- aurrea, on 10/12/2007, -1/+10You might be missing the word "Redundant" .... it's to protect your data. (it most cases)...
- surfing, on 10/12/2007, -0/+9Can't get to it either, but here's a good visual
Watercooler raid:
http://www.edvt.net/Pictures/raid.jpg - spectre, on 10/12/2007, -0/+9I take it somebody didn't Read The Article.
There are different configurations of RAID's and RAID 1, is intended to do exactly what you're insisting it can't, survive a node failure.
You're probably thinking of RAID 0, which has big problems on drive failures, because it is not redundant. - FTLJohnson, on 10/12/2007, -2/+9I had seen this a long time ago, but this is how I learned what raid was all about. Definitely the best tutorial out there.
- JoVoCop, on 10/12/2007, -3/+9I've heard it as
-Redundant Array of Independant Disks
-Redundant Array of Intelligent Disks
-Redundant Array of Inexpensive Disks
But I think 'Independant' is the most used. - SuperSloth, on 10/12/2007, -1/+7No, his argument is invalidated by the fact that servers still use SCSI RAID. Just like they always have. And probably still will for the next 5 years *at least*.
SATA is still a slower interface than the fastest SCSI (Ultra640 currently, going faster still pretty soon) and running a real SCSI RAID controller for RAID 5, 10 (1 0), 50, or 51 instead of the ***** CPU run chipset garbage they sell on enthusiast boards. You try running a database server on a system with a chipset RAID controller running both the OS, the SQL server, and the parity checksums through the CPU.
I'm laughing just thinking about it. - briansmith, on 10/12/2007, -0/+5Your RAID level isn't the only thing, or even the primarily thing, that determines your fault tolerance. You need to make sure the array is configured properly, and that you practice recovery from many different types of failures (controller failure, multiple disk failure, total power failure, etc.) I recently helped an admin who had a small RAID 10 array (4 pairs of mirrored disks, striped together, with two hot spares) that had a major failure. Due to either a hardware bug or some configuration mistakes, the automatic recovery did not successfully complete. Furthermore, when the admin tried to repair the problem manually, the whole array was broken to the point where some important data files were unrecoverable.
When something really bad happens, an array that utilizes striping will be more difficult when it comes time to do disaster recovery (pulling raw usable data off disks). The reason is that pieces of every file are split across many (4, 8, 12, or more) disks. Therefore, if you do not have a significant performance need for striping, you should avoid it. Similar arguments apply for RAID3-RAID5.
If a component has 99.99% reliability, then that means that its likeliness of failure is 0.01. But, if you have 10 of them, then the likelyness of failure is 10x0.01=0.1 (ten times as likely). So, implementing RAID actually increases the odds that you will experience a hard drive failure simply because you will be using more disks. To mitigate this, you should avoid using more disks than necessary to meet your goals. For example, if you need 144GB of storage, it is better from a reliability point of view to choose RAID1 with two 144GB disks than it is to choose RAID10 with four 72GB disks, assuming the two kinds of disks have equal RAS characteristics.
However, some applications require more disks (usually called
spindles) to meet their performance objectives (RAID10 with eight 72GB disks is generally faster than RAID1 with two 144GB disks, all things being equal).
The philosophy of RAID is that even though you are more likely to experience a hard drive failure (because you are using more disks), such a failure is much less likely to cause a loss of data/service due to redundancy.
Some people take for granted that software RAID is inferior to hardware RAID. But, that issue is a lot more complicated than it looks. Firstly, many (most) applications can be tuned so that the performance differences between software and hardware RAID are inconsquencial. Furthuremore, software RAID potentially provides more opportunities to detect hardware failures. See http://blogs.sun.com/roller/page/bonwick?entry=zfs_end_to_end_data (not my website) for an explanation of that.
The final thing (probably should have been the first thing) to note is that hardware failures are not the biggest danger to the data you are trying to protect. You (the administrator), or your users, are much, much more likely to cause a loss of data than the hardware is. With that in mind, a well-planned, well-tested, rock-solid backup and restore strategy should be a much higher priority than decisions regarding what kind of RAID to implement. I would bet anybody $1,000 right now that multiple people reading this would not be able to correctly restore their most mission-critical system from backup, even given unbounded time constraints. That is not good. - dhughes, on 10/12/2007, -0/+4 That site has been around for awhile, it's explanations are nice and clear and the design (still the same) is clean and easy to understand.
A 0+1 setup seems to be the best solution for home users, if you really need it is another question, but it would be fun to try.
It's down go to Wikipedia instead: http://en.wikipedia.org/wiki/Redundant_array_of_independent_disks#RAID_0.2B1 - master_of_fm, on 10/12/2007, -0/+4raid 0 has never been "fault tolerant" i imagine that was the reasoning behind deciding that striping was called RAID 0, because you have zero chance of recovery. an easy mnemonic device has always been to to associate RAID 0 and RAID 1 with binary, 0=no (not fault tolerant) 1=yes (fault tolerant)
- quastor, on 10/12/2007, -3/+7He's probably thinking of RAID 0, which isn't actually redundant.
- command-tab, on 10/12/2007, -0/+2Way to promote a product with a tutorial.
Get a RAID card and some SCSI drives off eBay and configure it. The software's not hard to use. - abyssknight, on 10/12/2007, -0/+2I believe you are right. It was originally called 'inexpensive' (as taught in my database systems class) but as of late is called 'independent' due to new technology. I think.
- wolphcry, on 10/12/2007, -1/+3Read about RAID level 5 again. I took 6 old 20gig hard drives and made a software RAID 5. Works great for storing digital photos that i don't want to lose. Just make sure you have an extra drive of the correct size to off to the side incase one dies.
- MikeSD34, on 10/12/2007, -1/+3aurrea has this right, the original purpose of RAID was that in the event of a hard drive failure, your data is already backed up, without you having to go through any periodic ritual of syncing or saving your data.
There is a RAID setup called stripping that increases data access speed, but that is less common ( and less stable ) than other forms, though when put together with the other forms ( RAID 0 1 ) can be very helpful. Read the 'article', you'll get it eventually.
Edit: I've been beaten to the punch, repetitively, but as long as the point gets across correctly, I'm happy. - master_of_fm, on 10/12/2007, -0/+23ware 9000 series controller with at least 5 WD 150GB raptors in RAID5 (4 active, one hot spare) in a 64-bit 133 slot with 2 dual-core opertons, on tyan motherboard sould be so awesome
- muldoonaz, on 10/12/2007, -3/+5@jumpertk: RAID 0 is Striping, not mirroring.
I use a combination of RAID Array's to achieve my goal with my various file servers both at home and at work. Striping is no good because if one drive fails, the entire array set is lost. However, RAID 1 0 gives you some fallover and time to replace failing drives if you plan everything out right.
Overall, I think RAID is a tad overrated for home use, but its great for redundant storage and high-access servers.
fyi, its always been Redundant Array of Inexpensive Disks. - Insimbi, on 10/12/2007, -1/+3digg effect :(
- spikes, on 10/12/2007, -0/+1Its arguable if mode 0 is even a real RAID mode. To me it isn't, its not RAID if there is no redundancy.
- cerealrat, on 10/12/2007, -0/+1http://www.acnc.com.nyud.net:8080/04_00.html
- whiteV6, on 10/12/2007, -0/+1I have to say for knowing nothing about raid, after reading and watch the diagrams i have a full under stadding for Raids now. Good +Digg
- jambarama, on 10/12/2007, -1/+1I think a RAID 5 is completely usable for home too. Especially if you have a hardware implementation. 3 disks in RAID 5 and you lose 33% capacity - in a raid 01 you lose 50%. If you have 5 drives you lose 20% in RAID 5, you still lose 50% in RAID 01 (or 10). As you know from reading the article, a RAID 5 is fault tolerant. If one drive goes down, you don't lose any data. Plus read/write speeds are much faster than RAID 01, plus you lose less capacity to redundancy.
- koweja, on 10/12/2007, -0/+0or independence for that matter
- sinembarg0, on 10/12/2007, -1/+1Your argument is negated by the fact that RAID no longer implies SCSI.
- jugo23, on 10/12/2007, -5/+5Was always confused about RAID, till now.
Good tutorial - merreborn, on 10/12/2007, -2/+2RAID 0 is the exception, not the rule. Every other RAID configuration I can think of is fault tolerant.
And yes, RAID 0 sucks. It halves your MTBF, instead of increasing it. It's best avoided entirely, except in very rare situations where you're dealing with large volumes of ephemeral data. - insaneses, on 10/12/2007, -0/+0How is this a 'Tutorial'? It just looks like a page explaining different levels of RAID.
I'm sure you could have found something better than just the first google result for 'RAID tutorial'. - crossers, on 07/04/2008, -0/+0very intersting tutorial. thanks!
http://www.ocflex.com/
http://www.trgovinca.org
http://www.chasr.org/ - MikeN, on 10/12/2007, -0/+0A very good tutorial! Whenever I hear noobs in the department ask "What's RAID?" they soon find a link to this site in their inbox. They're soon up-to-speed.
- caiser05, on 10/12/2007, -0/+0I like this tutorial
http://www.orbitmicro.com/support/resources/raid_tutorials.html - SuperSloth, on 10/12/2007, -4/+4No Digg. This is the good RAID tutorial:
http://www.pcguide.com/ref/hdd/perf/raid/index.htm - h198x, on 10/12/2007, -1/+1'Ever' in the title?
N O D I G G! - luckybano, on 10/10/2007, -0/+0thanks
http://www.w-wv.com - koweja, on 10/12/2007, -0/+0Excellent article. RAID is (or was) the latest trend in home computers, so you hear a lot of people and magazines talking about how you should use RAID, but the never really explain what it is. It's just "get two hard drives and set them up in a RAID 1 configuration".
- odin, on 10/12/2007, -2/+1it would be cool if it would load
- Platypus, on 10/12/2007, -1/+0If you're using RAID-5 you'd better have not only a hardware implementation but a *good* hardware implementation - preferably external. Software RAID-5 simply sucks donkey balls. You might think host-based controllers would be better, but I've tested a few and they really aren't. They can do RAID 0 or 1 OK, but not RAID 5. For that to be effective, you pretty much need a large non-volatile cache (such as external disk arrays have). This is necessary both to improve the chances of having the parity block already available and to complete the write early while parity calculations and back-end destaging are done asynchronously. Array vendors don't spend that much time designing such caches just for fun.
- tracer200, on 10/12/2007, -2/+1yes, it is Redundant Array of Inexpensive Disks
- rotten777, on 10/12/2007, -3/+1too bad the summary is somewhat misleading. the term "fault tolerant" means 1 member of an array can die without the array failing or losing data. this is not always the case as it explains in the article. a 2 drive RAID 0 array that has 1 disk fail is done for as is the data.
- levig, on 10/12/2007, -2/+0Opps... I did do a quick search, but I guess my URL was too specific... thanks for pointing it out.
- pixogen, on 10/12/2007, -3/+0Search next time: http://digg.com/search?search=http%3A%2F%2Fwww.acnc.com%2F&submit=Submit
- freestyle, on 10/12/2007, -7/+3wow the site is dying already!
- levig, on 10/12/2007, -4/+09 out of 10 isn't bad... :)
- jumpertk, on 10/12/2007, -8/+2Yeah.. that's why most servers that want to make sure their data has a real time backup in between daily, weekly, monthly backups onto tapes, DVDs or Cd's run RAID because "Raid is for killing data" RAID 0 is mirroring and yes if one disk goes bye bye you are kinda screwed but at least you got that speed performance for gaming :)
- elnerdo, on 10/12/2007, -14/+2Comment of the day!
- JaggedEdge, on 10/12/2007, -26/+2been there done that.... the whole RAID and SCSI thing is kinda outta "style" :P
- SamX, on 10/12/2007, -30/+1Raid is for killing data. u get the speed but in most RAIDs when one goes bad the others are useless.
Browsing Digg on your phone just got easier with our enhancements to the 
© Digg Inc. 2009
— Content created and posted by Digg users is 