[ The PC Guide | Systems and Components Reference Guide | Hard Disk Drives | Hard Disk Performance |
Hard Disk Quality and Reliability ]
Redundant Arrays of Inexpensive Disks (RAID)
One new technology that has evolved in recent years is the use of multiple disks in
conjunction, to improve performance and enhance reliability. This system is called redundant
arrays of inexpensive disks or RAID. RAID improves both performance and
reliability by storing data onto multiple physical disk drives. Reliability is improved
through redundancy of information, and performance is improved by exploiting the fact that
multiple disks are available for use. There is always a penalty when things sound this
good, of course, and that penalty is cost.
There are several standard implementations of RAID, called levels, as well as
many derivative or proprietary versions. Note that these are not hierarchical in terms of
quality, so level 2 RAID isn't necessarily "better" than level 1 RAID, just
different (the possible exception being level 0). A full discussion of RAID is beyond this
web site, but the following are summaries of the most common RAID levels:
- RAID 0 (Striping Without Parity): The lowest level of RAID isn't even really
RAID, because it does not offer redundancy. In this configuration, data is striped between
multiple drives. So if you were writing four sectors of data and you had three disks in
the array, sector 1 would go to drive 1, sector 2 to drive 2, sector 3 to drive 3, and
sector 4 back to drive 1, and so on, creating virtual "stripes" across the disk
array. Doing this greatly improves performance on both reads and writes due to
parallelism, but since there is no redundancy, there is no data protection.
- RAID 1 (Mirroring): In this configuration, usually used with two disks, every
write to the array is done to both disks. This means that there is an exact duplicate of
every piece of data in the system. This provides the highest data security because even if
an entire drive crashes, there is no data loss due to the duplicate. Many systems set up
this way will recover from a downed drive automatically without even a hiccup. The system
also greatly improves read performance; since there are two identical disks, any piece of
data required can be fetched from whichever disk is idle or whichever one has its head
assembly closest to the cylinder containing the data, for example. The drawbacks of RAID 1
are of course cost (you are doubling your storage requirements) and also write
performance, since every write must be done twice. Of course in most server environments,
reads outnumber writes by a large margin.
- RAID 5 (Striping with Parity): Probably the most popular type of RAID being used
in small to medium sized LANs today, RAID 5 is a good compromise between cost, performance
and reliability. RAID 5 works by striping across the disks, much the way RAID 0 does, but
it adds parity information that can be used to reconstruct any lost data in the event that
one of the disks in the array fails. This parity information is similar in concept to the parity and ECC codes used for system memory and within hard disk sectors themselves. This arrangement
gives it most of the data protection of RAID 1, with most of the performance advantages of
RAID 0. Instead of devoting 50% of your storage to redundant information as with RAID 1,
you devote only 25% when using a four-disk array. If any of the disks goes down, the
parity information stored on other disks can be used to reconstruct the lost data. It
isn't as safe as RAID 1 because you don't have the complete redundancy, but it is cheaper
to implement and provides theoretically better overall read performance, as well as better
write performance.
Implementing RAID can be done either in hardware, or in software. Hardware RAID is
normally done either by a dedicated RAID controller, or by a SCSI host adapter. SCSI host
adapters vary widely in terms of their ability to support various RAID levels, so you need
to determine what each can support on an individual basis. RAID can also be implemented in
software, by the operating system. One operating system that supports RAID internally is
Windows NT. Generally speaking, hardware RAID is superior because it is faster.
Next: Viruses and Virus Protection
Home - Search
- Topics - Up