RAID Disc System
Another important issue for discussion in determining the specifications -
server redundancy. There are several methods of improving the reliability of
disk systems from multiple disks. Most of these schemes redundancy - RAID
variation (stands for, or redundant array of inexpensive disks independent;).
The original specification RAID has been designed to replace the large and
expensive disks mainframe minicomputers and arrays of small and cheap discs
designed for minicomputers, hence the word inexpensive. "Unfortunately, the
systems RAID rare something inexpensive. RAID implementations is a series of
redundant disk arrays for provide different levels of protection and the speed
of data transmission. As involves the use of RAID disk arrays, the best the
application will interface to SCSI, because it can support up to 15 devices.
RAID Levels there 6: from zero to fifth. While some manufacturers advertise
their own scheme redundancy, which they called RAID-6, RAID-7 or above. (RAID-2
and RAID-4 is not on the network servers, so we are not going to talk.) Of all
levels of RAID zero is the highest productivity and lowest security. It implies
the existence of at least two devices, and synchronized recording data on both
disks, the disks look like one physical device. The process of recording data at
multiple disks called filling disks (drive spanning), a proper method of the
data-recording their alternating (data striping). When sequencing data are
written at all poblochno disk, a process called delamination blocks (block
interleaving). The block size is determined by the operating system, but it
usually ranging from 2 KB to 64 KB. Depending on the design of disk controller
and HBA, the sequential operation may overlap, resulting in productivity
increases. Thus, in itself RAID-0 can increase productivity, but does not
provide protection from the bug. If a crash occurs, so goes the entire subsystem
down, which usually leads to a complete loss of data. option is alternating data
distribution data (data scattering). As with penetration, data are recorded
consistently for a few discs filled. But unlike the sandwich record is not
necessarily done on all disks, or if the drive is busy, data can be recorded on
the next drive is accessible lets you add drives to existing vols. As standard
RAID-0, the combination fill discs with the distribution of data improves
productivity and increases the volume, but does not provide protection against
failures RAID-1, also known as mirroring disk (disk mirroring), involves the
installation of pairs of identical drives, each drive in a pair is a mirror of
another. In RAID-1 data are written into two identical or nearly identical pairs
of disks: when, for example, a disc degenerates, the system continues to work
with a mirror disk. If the mirror disks are common HBA, the performance of this
configuration, compared with odnodiskovoy, will be less because the data should
be recorded in sequence on each disc.
Next part
|