Promise FastTrack/FastSwap IDE RAID Controller

If you have an important document that you simply must take care of, what is the first precaution that you would normally take to insure its safety? You would make sure you had a backup copy of course, so that if anything were to happen to the original, you would have a complete backup ready and waiting for immediate usage. This same principle is mimicked in the next level of RAID, known as mirroring (RAID 1).

When a disk array is setup in a RAID 1 system, each member of the subsystem contains a perfect copy of each other, and when a single file is written to the disk (which appears to be a single drive, equal in capacity to that of the smallest drive in the array multiplied by the number of drives and finally divided by two for redundancy) it is written to all drives in the array. Should a single drive in the array fail, the other drive or drives will act as a replacement, often resulting in zero downtime (or very little) for the mission critical system that happens to be running with the failing drive. This technology is also known as fault-tolerance.

Note that the method for calculating the total usable disk space for a mirrored drive array is different from that of a RAID 0 system in that the total capacity is only equal to the capacity of the smallest drive. This is because all mirrored drives act as just that, mirrors, and cannot be treated as independent drives once in the RAID 1 setup for data storage. The only time the mirrors activate is when the primary drive fails, at which point a mirror will come online and replace the primary drive.

Mirroring is a technology that is utilized in many servers where a single drive failure would result in an unacceptable amount of downtime, and in the Internet business community, any downtime is unacceptable.

As you might be able to guess, because of the fact that every file must be written to all members of the RAID 1 array, the write performance of a RAID 1 system is considerably slower than a single disk, and definitely slower than a RAID 0 system. In contrast, the read performance is considerably faster since data can be read, just like with a striped array, in parallel from all members of the array.