The Schworak Site | Log In | Up One Level

RAID - What is it?

RAID - Redundant Array of Independent Disks, which was originally Redundant Array of Inexpensive Disks. Basically, this is a way to connect multiple physical disks together to create one logical disk. Some versions of RAID offer fault protection and others to not.

RAID 0

RAID 0 requires at least 2 drives. All of the available drive space is used for storing data.

RAID 0 has no fault protection. If any one drive fails, it is likely that all data will be lost and you will need to restore from a backup.

It is best used for systems that need to merge multiple disks together to gain drive space as if it is a single larger drive. It also offers increased speed because any one file can be read from multiple drives at once with each drive providing part of the overall file.

A good example might be if you were streaming video. You could keep your original video stored in a backup location that doesn't get used but when people stream it each read more quickly by taking part off each drive.

A general file server would be another good example assuming it was mainly for reading.

If you are going to be writing important information to this file system, you will need to back it up frequently or you risk losing it if a drive crashes.

RAID 1

RAID 1 requires at least 2 drives and must be expanded in multiples of 2 drives. Half of the drive space in the set is used for data and the other half is used as a backup or mirror of the data.

RAID 1 has fairly good fault protection. If any one drive of a pair dies, the data on the other drive of the pair is in tact.

This is a very stable and popular RIAD configuration. It is as fast as RAID 0 for writing but not quite as fast for reading because all reads come off the master drive alone.

This is also known as a MIRROR configuration. That is because the data on the backup drive is an exact copy of the data on the master drive.

RAID 2 / RAID 3 / RAID 4

RAID 2, RAID 3 and RAID 4 are basically all the same. The difference is how small the files are diced up. RAID 2 splits files at the bit level, RAID 3 splits at the byte level and RAID 4 splits at the block level. In each case, you will need at least 3 drives. The parity in each case is stored on one deticated drive so all but one drive are used to store your data.

These formats offer fair fault protection. If any one drive is lost, the information on the missing/damaged drive can be calculated using the remaining drives. If a second drive is lost before the first drive is replaced, data will be lost.

These formats are not used very often any longer due to the nature of the deticated parity drive. It causes a bottleneck when trying to read data off of the drive array.

RAID 5

RAID 5 is similar one of the most popular configurations because it only uses a single parity stripe but it can put that stripe on any of the available drives. A minimum of 3 drives are needed for this configuration like RAID 2, 3 and 4 but because the parity stripe is not on a deticated rive, reading is much faster because there is less of a bottleneck.

As with RAID 2, 3 and 4, any one drive can fail without any data loss. The information on the failed drive can be calculated from the information on the other drives. If a second drive fails before the first drive has been replaced and rebuild, data will be lost.

The amount of drive space is always n-1 which is the total number of drives less one. More drives offers faster read times. Write times generally don't improve with added drives.

RAID 6

RAID 6 is actually a RAID 5 drive with a second parity stripe added. A minimum of 4 drives are needed for this configuration.

Fault protection is better than RAID 5 because if one drive fails and then a second drive fails before the first drive is replaced and recovered, no data is lost. Because it is unlikely that 3 drives would fail at once this configuration is very stable, especially in arrays with a large number of drives (more than 6).

The main draw back to RAID 6 is the slower write speed. Each parity stripe is calculated independantly thus RAID 6 takes a little more than twice what RAID 5 does for the parity calculations. The extra time is due to the difference in the calculation methods used.

When a RAID 6 drive fails, the array falls back to RAID 5 mode and keeps running. It is still considered to be in a degraded state and you should replace the failed drive as quickly as possible.

As a basic rule, you shouldn't consider RAID 6 if you are running less than 5 disks because of the overhead in space. When running 5-6 drives it is a bit of a toss up and really depends on your desire to keep your data safe. But  when running more than 6 drives you really should be running RAID 6 because the odds of losing more than 1 drive at a time start to go up. Especially if you put multiple drives into service at the same time.

RAID 01 aka RAID 0+1

These are actually stacked algorythms where two RAID 0 arrays are created as and then they are combined into a RAID 1 array.  As with RAID 1, 50% of the available drive space is used as an on the fly backup or mirror. There must be multiples of 2 disks with a minimum of 4 to start the array out. This combines the speed advantates of RIAD 0 but also gives the redundancy advantage of RAID 1.

RAID 10 aka RAID 1+0

This is similar to RAID 01. The difference is that the RAID 1 is created first and then the RAID 0 is stacked on top of that. You get the same speed and redundancy as with RAID 01. In theory, RAID 10 offers a bit more fault tollerance and much better recovery times when compared to RAID 01. Given the choice, I would recommend RAID 10 over RAID 01. 

Which RAID you use depends on your needs. You will likely find that if you only have a couple drives and you want to prevent data loss in the event of a failure, RAID 1 is your best bet. You can boot your system from a RAID 1 configuration and if you are really concerned about your data you can even run RAID 1 with 3 drives. The third drive will be a "spare" and won't be used unless one of the other drives fails. In that event the system will automatically start using the spare drive to keep your data safe.

The concept of using a "spare" drive actually works with with RAID 1, 5 and 6. It may also work with 2, 3 and 4 but having never actually used these formats myself I can't say for sure. You cannot use a spare on a RIAD 0 configuration because there is no fault protection for RAID 0. It is possible to use spaces on the RAID 1 portion of a RAID 10 array but you would have to dedicate an entire array of disks as the spare in a RAID 01 configuration so that is not recommended.

RAID 5 is currently the most popular configuration for systems with 3-5 drives in the array (and larger in some cases) where RAID 6 is gaining popularity in arrays with 6 or more drives. As mentioned earlier, it is some what of a toss-up on arrays with 5-6 drives as to which way you should go. It really depends on what is more important to you, write speed or fault protection.

Comments

<< Score an A+ on SSL Labs using HAPROXY    |    Google Chrome ~ Disable Swipe to go Back or Forward >>
All content on this site is copyright ©2004-2024 and is not to be reproduced without prior permission.