What the heck is this "Raid 5" I've been hearing about? Raid 0 is data striping and Raid 1 is data mirroring............
Where did raid 5 come from?
Where did raid 5 come from?
Raid FIVE?
- Thread starter The Pentium Guy
- Start date
Gamingphreek
Lifer
- Mar 31, 2003
- 11,679
- 0
- 81
Raid 5 is the same as RAID 0 only it has Fault Tolerance. So the speed of RAID-0 with the transfer rates and data protection (fault tolerance) of RAID 1. THe only stipulation is you need 3 HDD's in order to use it.
IIRC there is a RAID 7, 8, 9, 10, and 50.
Cant remember what they all do off the top of my head but RAID 10 is a larger more powerful version of RAID 5 IIRC.
-Kevin
IIRC there is a RAID 7, 8, 9, 10, and 50.
Cant remember what they all do off the top of my head but RAID 10 is a larger more powerful version of RAID 5 IIRC.
-Kevin
Raid 3 and Raid 5 use a combination of striping and mirroring to improve data reliability, maximize storage, and improve drive array performance. These types of RAID arrays are usually found in SCSI-based server storage solutions and usually involve a minimum of three or more hard drives.
Essentially, the data is divided into multiple parts and then each of those parts is written to multiple hard drives. For example, data parts A, B, and C might be written to hard drives 1, 2, and 3. Drive 1 holds parts A and B, drive 2 holds parts B and C, and drive 3 holds parts A and C.
If any of the three hard drives fails, all segments of the data are still available. While running normally, any segment of the data can be pulled from multiple drives simultaneously. Storage is improved vs. RAID 1 (mirroring) as well. In RAID 1, two 100Mb hard drives provide 100Mb of storage. In RAID 3/5, 3 100Mb drives provide 200Mb of storage. If two drives fail, you're screwed.
There is a technical difference between how RAID 3 accomplishes this versus RAID 5, but off the top of my head I don't remember.
There is also the ability to add a "hot spare" to an array, which is essentially a stand-by drive that waits for an active drive in the array to fail. When an array drive fails, the RAID controller rebuilds it onto the hot spare with data pulled from the surviving disk drives. The array can then continue running with redundancy and the failed drive can be removed at the convenience of the server administrator. A file server or storage system with "hot swappable" drives can have the bad drive removed without shutting down the file server for maintenence.
Rebuilding an array on the fly can impact overall system performance and may not be desirable on a server or storage system under heavy load. In high load environments, RAID 3 and 5 arrays can be clustered so a backup array can take over while a failed drive in an array is replaced and the array rebuilt. Huge databases can also be striped across multiple RAID arrays for enhanced performance. You may see references to RAID 30/50 (31/51) or RAID 3+1/5+1 (3+0/5+0).
These types of sophisticated storage systems generally involve multi-channel SCSI RAID controllers and very expensive hard drives.
Hopefully this answers your question.
Essentially, the data is divided into multiple parts and then each of those parts is written to multiple hard drives. For example, data parts A, B, and C might be written to hard drives 1, 2, and 3. Drive 1 holds parts A and B, drive 2 holds parts B and C, and drive 3 holds parts A and C.
If any of the three hard drives fails, all segments of the data are still available. While running normally, any segment of the data can be pulled from multiple drives simultaneously. Storage is improved vs. RAID 1 (mirroring) as well. In RAID 1, two 100Mb hard drives provide 100Mb of storage. In RAID 3/5, 3 100Mb drives provide 200Mb of storage. If two drives fail, you're screwed.
There is a technical difference between how RAID 3 accomplishes this versus RAID 5, but off the top of my head I don't remember.
There is also the ability to add a "hot spare" to an array, which is essentially a stand-by drive that waits for an active drive in the array to fail. When an array drive fails, the RAID controller rebuilds it onto the hot spare with data pulled from the surviving disk drives. The array can then continue running with redundancy and the failed drive can be removed at the convenience of the server administrator. A file server or storage system with "hot swappable" drives can have the bad drive removed without shutting down the file server for maintenence.
Rebuilding an array on the fly can impact overall system performance and may not be desirable on a server or storage system under heavy load. In high load environments, RAID 3 and 5 arrays can be clustered so a backup array can take over while a failed drive in an array is replaced and the array rebuilt. Huge databases can also be striped across multiple RAID arrays for enhanced performance. You may see references to RAID 30/50 (31/51) or RAID 3+1/5+1 (3+0/5+0).
These types of sophisticated storage systems generally involve multi-channel SCSI RAID controllers and very expensive hard drives.
Hopefully this answers your question.
So what's the difference between raid 0+1 and raid 5? I guess 0+1 uses 4 drives, can't repair on the fly, and splits data differently?
raid0 = 2 drives minimum, striping data in parts, read and write speed increased (theoretically doubled, speaking from experience +88% 
), space=smallest drive*#drives
raid1 = 2 drives minimum, data written identically to two drives, read speed increased, write speed not (same as 1 drive), space=smallest drive*#drives/2
raid 0+1 = 4 drives minimum, combines above, read and write speed increased, space = raid1
raid5 = 3 drives minimum, data split into parts PLUS redo-information to rebuild lost data in case of failure, read speed increased, write speed normal (data split up, but also redo-log has to be written), space = (smallest drive*#drives)*(#drives-1)/#drives, e.g. 3*80GB in raid5=160GB array
), space=smallest drive*#drivesraid1 = 2 drives minimum, data written identically to two drives, read speed increased, write speed not (same as 1 drive), space=smallest drive*#drives/2
raid 0+1 = 4 drives minimum, combines above, read and write speed increased, space = raid1
raid5 = 3 drives minimum, data split into parts PLUS redo-information to rebuild lost data in case of failure, read speed increased, write speed normal (data split up, but also redo-log has to be written), space = (smallest drive*#drives)*(#drives-1)/#drives, e.g. 3*80GB in raid5=160GB array
Raid 0+1 will be slightly slower than Raid 5. Only parity is written to the extra or odd numbered drive in raid 5; whereas in Raid 0+1; the mirrored drive(s) gets all the data; therefore a more complete safety package, but a few nano secs slower !!
If a raid 0+1 has a drive failure, can it be repaired?
I realize will should still work, there will be two drives with good info, but what about the drive set that is no longer working, how could you get it to be mirroring the first pair again?
I realize will should still work, there will be two drives with good info, but what about the drive set that is no longer working, how could you get it to be mirroring the first pair again?
Yes.... Raid 0+1 = 2 drives raid 0 striping + Raid 1; 2 drives mirrored; if either of a raid 0 drive fails, replace the drive and cross write from the raid 1 drives. If a raid 1 drive, fails again you replace the drive and cross write from the other good mirrored drive. If both raid 1 drives were to fail you are SOL.
To complete the answer ... the Raid controller will know what data needs to be added to the new drive ?? More than likely it just rewrites the good data from the mirrored to both the Raid 0 drives. Over-writing the one drive that didn't fail.
I have used both raid 0 and Raid 1. I never used the Raid 0+1 system. After I lost a drive in Raid 0 and didn't have the data backed up; I never used it again. I now use Raid 1 on PC1 & PC2. I can afford 2 HD's per system but not 4 !. I'm not a gamer; so I don't need the slight increase in speed that raid 0 offers.; but I don't like losing my data. I go a step further by having an external HD on each system that I send a backup of the system to, scheduled every 2 weeks.
Maybe Raid 0 plus an external HD that gets the system backup would be a cheaper way to go and still have your data protected.
I have used both raid 0 and Raid 1. I never used the Raid 0+1 system. After I lost a drive in Raid 0 and didn't have the data backed up; I never used it again. I now use Raid 1 on PC1 & PC2. I can afford 2 HD's per system but not 4 !. I'm not a gamer; so I don't need the slight increase in speed that raid 0 offers.; but I don't like losing my data. I go a step further by having an external HD on each system that I send a backup of the system to, scheduled every 2 weeks.
Maybe Raid 0 plus an external HD that gets the system backup would be a cheaper way to go and still have your data protected.
See sig
I really appreciate the speed kick of RAID0 (especially with the nice SATA-II drives) and have regular backups of important data and images of the boot partition ready on the "slow" drive...LOL...that's really all I have to say. Do a simple lookup on google and you can find a ton explaining what the all the raid levels mean.
dhoytw: If we all used google; and could find the answers ... would there be a need for AT forums ... or YOU.
Be warned that if you use a cheapo controller for RAID 5 that is software based it is going to take a LOT of CPU to run it, and will likely be very slow. That is why most RAID 5 controllers are expensive, hardware based with processor on the card. Also notice most good solutions also require a motherboard with 64 bit PCI-X slots also. How about something like this in your PC with a good raid 5 controller, it does SCSI too. As Homer Simpson would say, Uhhhhhhh!! It might not compare to my Sunny 4CPU server at work, but it would be fun to play with.
http://www.pcstats.com/articleview.cfm?articleid=1641&page=2
http://www.pcstats.com/articleview.cfm?articleid=1641&page=2
Although najames is mostly correct, good RAID 5 cards have been available for many years on the PCI-32 and then the PCI-64 card form factors. I've been running RAID 5 on at least one of my PCs for several years now. Older hardware, but on the server it makes very little difference (to me.)
The first time you have a drive failure, and the only thing you do is pull out a drive, and push one in, can give you a real high. Especially if you've ever tried to recover data after a HD failure w/o RAID!
Jeff7
Lifer
- Jan 4, 2001
- 41,596
- 19
- 81
Promise has some cards that run on a 33MHz PCI slot. They just do better if they've got a faster PCI slot to play with. I'm running an SX4000 card with 4 drives on it - excellent speeds and low CPU utilization. Their SX4060 has 64MB of RAM soldered to the board, and their S150 SX4(-M) cards run SATA drives (the -M has 64MB of RAM on it).
It occurs to me...I should have mentioned that I was talking about all-SCSI systems. No IDE, no ATA, no SATA.
"IIRC there is a RAID 7, 8, 9, 10, and 50. "
RAID 0 is block-level striping Two drives, minimum.
RAID 1 is mirroring. Two drives, minimum.
RAID 2 is bit-level striping with ECC. It's not used anymore. It required _14_ disks, and spindle-locked moves. See why no one uses it?
RAID 3 is an array of disks with a single parity disk, striped by byte. Three drives, minimum.
RAID 4 is RAID 3 with block-level striping.
RAID 5 is RAID 4 with distributed parity.
RAID 6 is double parity RAID 5. (allows you to lose two drives)
RAID 7 is a proprietary standard similar to RAID 3-6. Only SCC uses it.
There's no such thing as RAID 8-9.
RAID 0+1 is striped mirrors (IE, you have two RAID 1s and then stripe them into a single RAID 0). RAID 10 is the opposite (you take striped arrays and mirror them). RAID 10 is faster than RAID 0+1. Neither is really a proper RAID level, since they're hybrid arrays. 4 drives, minimum.
RAID 50 is striped RAID 5 arrays. Again, not a real RAID level, just a hybrid. You could theoretically have RAID 0+5, too, except that there's no point (lower performance). 6 drives, miniumum.
You can actually build interesting hybrid arrays on a standard PC now, due to the influx of cheap 4-bay Firewire enclosures onto the Internet. If you're not worried about bandwidth, you could conceivably chain 64+ hard drives together, allowing for such oddities as RAID 51 (mirrored parity - you can lose up to half plus one of your disks!).
-Erwos
RAID 0 is block-level striping Two drives, minimum.
RAID 1 is mirroring. Two drives, minimum.
RAID 2 is bit-level striping with ECC. It's not used anymore. It required _14_ disks, and spindle-locked moves. See why no one uses it?
RAID 3 is an array of disks with a single parity disk, striped by byte. Three drives, minimum.
RAID 4 is RAID 3 with block-level striping.
RAID 5 is RAID 4 with distributed parity.
RAID 6 is double parity RAID 5. (allows you to lose two drives)
RAID 7 is a proprietary standard similar to RAID 3-6. Only SCC uses it.
There's no such thing as RAID 8-9.
RAID 0+1 is striped mirrors (IE, you have two RAID 1s and then stripe them into a single RAID 0). RAID 10 is the opposite (you take striped arrays and mirror them). RAID 10 is faster than RAID 0+1. Neither is really a proper RAID level, since they're hybrid arrays. 4 drives, minimum.
RAID 50 is striped RAID 5 arrays. Again, not a real RAID level, just a hybrid. You could theoretically have RAID 0+5, too, except that there's no point (lower performance). 6 drives, miniumum.
You can actually build interesting hybrid arrays on a standard PC now, due to the influx of cheap 4-bay Firewire enclosures onto the Internet. If you're not worried about bandwidth, you could conceivably chain 64+ hard drives together, allowing for such oddities as RAID 51 (mirrored parity - you can lose up to half plus one of your disks!).
-Erwos
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter