RAID Technology For Digital Video

RAID is an acronym for redundant array of independent (or inexpensive) drives. RAID technology concerns the use of storing and retrieving data on an array of multiple hard disks as opposed to a single hard drive. An array of RAID disks always has a controller built-in; the computer just ‘thinks’ it’s talking to a normal disk drive. All the ‘clever bit’ goes on inside the RAID array control device. Why use RAID? There are two reasons. The first is speed. Multiple disks, accessed in parallel, give greater data throughput (write/read speed) than a single disk. The second reason is reliability. With a single hard disk, you cannot protect yourself against catastrophic disk failure. Anyone who has experienced a total disk-drive crash will know the agony of installing a new drive; reinstalling the operating system and restoring files from backup tapes (assuming you’ve been careful enough to make these!) With an appropriate RAID disk array, your system can stay up and running when a disk fails. Moreover, RAID controllers are intelligent enough that, should one disk fail and be replaced with a virgin drive, it will rebuild the original array.

To some extent these two objectives are contradictory, and the term RAID covers several different arrangements, each with a different emphasis on speed versus reliability. Originally RAID came in five different varieties, termed RAID 1 to RAID 5. Some proved more useful than others. Recently RAID definitions have been extended (corrupted?), so you will sometimes see references to RAID 0, which is a scheme that uses multiple disks with no redundancy (and is therefore actually not RAID at all, but AID!). Similarly, you may see references to RAID 35, which is a mixture of RAID 3 and 5. Each of the five original RAID schemes is described below.

RAID 1 (mirroring) is usually called ‘mirroring’, and its emphasis is on data security. All disks in the array are arranged in pairs, and RAID 1 provides complete redundancy by writing identical copies of all data on these pairs of disks. For all its ‘belt and braces’ approach, RAID 1 still offers some increase in speed because writing to the disks can be done in parallel, whereas reads can be interleaved.

RAID 2 (bit striping with error correction) Unlike parallel RAID 1, RAID 2 works in series. The controller writes sequential blocks of data across multiple disks. Each sequential block is termed a stripe, and the size of the block is termed the stripe width. In RAID 2, the stripe width is 1 bit only. A RAID 2 system would therefore have as many data disks as the word size of the computer, and every disk access must involve every disk. In addition, RAID 2 requires the use of extra disks to store error-correction codes for redundancy. With 32 data disks, and a few parity disks thrown-in for good measure, it’s not surprising that RAID 2 has never been considered a practical option.

RAID 3 (bit striping with parity) is very similar to RAID 2, except that only one extra disk is used to store simple parity data. This parity disk is written with data derived quickly and simply from the 8, 16 or 32 data bits on the other drives. This only works because the disk controller of the drive which experiences the missing bit is able to report that it has had a data read error. Knowing which disk’s data is missing, the RAID controller can reconstruct the original data. For instance, imagine we write the byte 10010001 to eight RAID 3 drives. Assuming we use a simple even-parity scheme, we would write 0 as the data on the ninth parity drive because there is an odd number of ones in the original byte. So we would actually write the following across all nine drives:

1 0 0 1 0 0 0 1 (0)

where the (0) is the value on the parity drive.
Now, suppose in a subsequent read command we receive the following:

1 0 0  0 0 0 1 (0)

Digital video production 203
We have an even number of ones, but, because parity is 0 (or NOT EVEN), we know that the failed bit must have been a one.
The one drawback of RAID 3 is that it must read all data disks for every read operation. This works best on a single-tasking system with large sequential data requirements – for example, a broadcast quality videoediting system, where huge video files must be read sequentially.

RAID 4 (striping with fixed parity) is the same as RAID 3, except that the stripe widths are much greater; the intention being that individual read requests can be fulfilled from a single disk. However, this isn’t the case, because each read and write request has to access the single parity disk. This is such a drawback that RAID 4 is never implemented.

RAID 5 (striping with striped parity) uses large stripe widths and also stripes the parity across all disks. This scheme provides all the advantages of RAID 4, and it avoids the bottleneck of a single parity disk.

Related Products