Contemporary disc systems of RAID
The emergence and development of the concept of RAID related to the increase in
the use of data, the need to ensure the most rapid access and high reliability
storage. The amount of stored information has been calculated by many hundreds
of terabytes. It is now essential tool storage, providing quick access to
information is on magnetic disk drives, characteristics which in recent years
has grown. However, in comparison with purely electronic devices (CPU, RAM,
controllers, etc.) because of productivity growth electromechanical nature
drives markedly lower. Since the beginning of 80 years, the central processor
speed increased at an average of five, and by some estimates, even tenfold every
three years. External drive capacity doubled, falling in price by half to about
every three years. However, the access time for the period improved according to
different estimates by only 10-50%. As a result, external drives have become "bottleneck
".
Alas, soon no innovative solutions to realize the external drive without the
shortcomings of magnetic disks. For example, optical storage, capable of
providing much more specific capacity, suffer the same drawback: the
availability elektro-mehanicheskih knots. The flow of user queries very unevenly
distributed across the disk memory. The formation of the so-called "hot spots" -
areas to which the vast majority of treatment takes place, while the rest of the
memory is very low load. This situation applies known "rule 80/20" in most cases,
80% of hits implemented to a limited area that constitutes only 20% of capacity
memory. If this "hot spot" located on the same disk, external memory subsystem
performance will be very low, because the consecutive requests for most of the
time will be expected to end operations addition of the ball and finding the
desired sector on the disk. One of the approaches proposed for reducing the
adverse effects of rule 80/20, "is to use a more expensive device with a
significantly better performance for storage" hot files ". However, few in a
position to determine the likely nabolee set such files, the more so that,
depending on the type of tasks, the time of day, day of week and other factors
will inevitably change the picture. The most common solution to this problem is
a logical allocation of disk space to recruit (array) of relatively small and
relatively inexpensive disk devices with an average search time. "A hot spot"
located in the small fragments of the drive, made more uniform than their load.
With the apparent attractiveness of such a solution, and it has a very serious
flaw: the increasing number of disks in the array, and proportionally increases
the likelihood of rejection. In other words, in parallel with the increasing
number of working time developments in the drive waiver system as a whole is
declining, its reliability has become unacceptably low. Productivity growth in
the number of the drive will increase reliability and decrease. Increase the
reliability possible by the introduction of redundant capacity, providing an
opportunity to restore the destroyed data. RAID levels.
Next part
|
O
wlCom Software
