If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Display Modes |
#1
|
|||
|
|||
Beginner's question on RAID 1+0
What is RAID 1+0? A mirror of 2 stripe sets? Or a a stripe assembled
from lots of 2-disk mirrors? I've heard that one of those ways is bad, and thus the widely used RAID 1+0 uses the second way. Why the first way is bad? -- Maxim Shatskih, Windows DDK MVP StorageCraft Corporation http://www.storagecraft.com |
#2
|
|||
|
|||
Maxim S. Shatskih wrote:
What is RAID 1+0? A mirror of 2 stripe sets? Or a a stripe assembled from lots of 2-disk mirrors? I've heard that one of those ways is bad, and thus the widely used RAID 1+0 uses the second way. Why the first way is bad? If you take two RAID-0 stripe sets and make them mirror each other, the loss of a single disk in each set results in loss of data (and if the sets get large enough, the probability of that occurring is not all that low). When you instead mirror disk pairs and string the pairs together in a RAID-0 configuration, you only lose data if both disks in a single pair fail (the probability of this being far lower than that of the first case with any significant number of pairs - in fact, the ratio of data-loss probabilities may be about equal to the number of pairs, 2 pairs giving you about half the probability of data loss, 3 pairs giving you about 1/3, etc.). - bill |
#3
|
|||
|
|||
On Wed, 26 Jan 2005 02:50:20 +0300, "Maxim S. Shatskih"
wrote: What is RAID 1+0? A mirror of 2 stripe sets? Or a a stripe assembled from lots of 2-disk mirrors? I've heard that one of those ways is bad, and thus the widely used RAID 1+0 uses the second way. Why the first way is bad? 0+1 isn;t "bad" per se, it's just been ousted by a newer and, imo, better option 1+0. You are correct in your understanding of the layout. it's the second one you presented. The reason 0+1 is now considered bad is because the loss of a single drive causes the entire mirror'd plex to fall off, leaving you with a raid 0 plex of X drives. At the time this was the top of the heap for performance and protection. But enter 1+0..... This scenario allows you to lose up to half the drives in the raid set as long as 2 of them are not the same mirror'd plex. Illustration is generally better. stripe 1-1 mirror 1 2-2 mirror 2 3-3 mirror 3 The mirror's are the base plex's, then those plex's are striped. This scenario allows you to lose the X drives and still keep running. Basically half as long as they are the "right" half. stripe X-1 mirror 1 2-X mirror 2 X-3 mirror 3 Most array vendors and LVM software will do 1+0 now but that wasn't always true. So 0+1 was your best bet, particularly if you were using software. Hope this helps. ~F |
#4
|
|||
|
|||
If you take two RAID-0 stripe sets and make them mirror each other, the
Thanks! Is there any performance issues in this choice? -- Maxim Shatskih, Windows DDK MVP StorageCraft Corporation http://www.storagecraft.com |
#5
|
|||
|
|||
Maxim S. Shatskih wrote:
If you take two RAID-0 stripe sets and make them mirror each other, the Thanks! Is there any performance issues in this choice? Yes, and Faeandar sort of alluded to them. If you make two conventional RAID-0 stripes and then mirror them, then if you lose a disk in one of them that entire stripe may become unavailable (the software *could* be suave enough to use the remaining disks in the sick RAID-0 set for read requests that didn't require data on the failed member, but I wouldn't bet on it). This means that the disks in the remaining RAID-0 set are now shouldering the entire workload if just one disk in the other set fails. With mirrored disk pairs, the only performance impact when a single disk fails is that all read requests which would normally have been split between it and its partner now fall on the partner alone. But since that's only 1/Nth of the total request load (for N disk pairs), you've more or less still got 2N - 1 disks sharing the total load rather than just N (though the partner of the failed disk can in some cases still become a hot spot and drag down overall throughput more than that - e.g., if the data is striped so finely over the disk pairs that most read requests hit most of the pairs rather than having each read normally target only one or two pairs). - bill |
#6
|
|||
|
|||
RAID 1+0 is a mirroring of 2 stripes sets without a parity disk - this
provides awesome performance with almost no redundancy. It is also expensive as you require 50% of your disk space to be mirrored - yikes! "Maxim S. Shatskih" wrote in message ... What is RAID 1+0? A mirror of 2 stripe sets? Or a a stripe assembled from lots of 2-disk mirrors? I've heard that one of those ways is bad, and thus the widely used RAID 1+0 uses the second way. Why the first way is bad? -- Maxim Shatskih, Windows DDK MVP StorageCraft Corporation http://www.storagecraft.com |
#7
|
|||
|
|||
Monte Oates wrote:
RAID 1+0 is a mirroring of 2 stripes sets without a parity disk - this provides awesome performance with almost no redundancy. It is also expensive as you require 50% of your disk space to be mirrored - yikes! That doesn't make _any_ sense at all. RAID 1 (with or without +0) provides _full_ redundancy. Google is your friend... Scott |
#8
|
|||
|
|||
There's a huge difference in reliability between striping mirrored sets vs mirroring striped sets. You ALWAYS want to stripe mirrored sets. What this gives you in reliability is the best of all worlds - way better than standard RAID-5 in fact. The reason is that the probability of losing both drives in a two-disk mirror set in the same time window is very small. However, if you stripe mirror sets, losing a single drive loses and entire 1/2 of the whole mirror set, and as the number of drives in the entire configuration grows, the probability of losing two drives at the same time grows. Does this short explanation help at all? On 22 Feb 2005 08:53:09 GMT, Scott Howard wrote: Monte Oates wrote: RAID 1+0 is a mirroring of 2 stripes sets without a parity disk - this provides awesome performance with almost no redundancy. It is also expensive as you require 50% of your disk space to be mirrored - yikes! That doesn't make _any_ sense at all. RAID 1 (with or without +0) provides _full_ redundancy. Google is your friend... Scott --- jls The preceding message was personal opinion only. I do not speak in any authorized capacity for anyone, and certainly not my employer. (get rid of the xxxz in my address to e-mail) |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
How I built a 2.8TB RAID storage array | Yeechang Lee | Homebuilt PC's | 31 | February 22nd 05 06:40 PM |
Another RAID for Beginners Question | Philadelphia Frank | Asus Motherboards | 4 | May 31st 04 01:59 PM |
SATA RAID question... basic RAID question | Scotter | Asus Motherboards | 9 | January 2nd 04 11:08 PM |
DAW & Windows XP RAID Tips, ProTools error -9086 | Giganews | Asus Motherboards | 0 | October 24th 03 06:45 AM |
RAID and non-RAID combination | Howard | Gigabyte Motherboards | 3 | October 4th 03 11:54 AM |