Contemporary disc systems of RAID
RAID-1. The technology, which is based on the configuration of RAID-1, known for
long and named "mirror disks." As before, a user array is a virtual disk device
with the necessary amount of memory, broken into segments. These are placed in
the same way as traditional disk tonnes. E. segments A, B, C, ... located in the
natural order of the first disk. When the disk is full, the data posted in the
same way on the third drive, and so on until exhaustion of total memory.
Redundancy is achieved by the second disk, which is the exact copy of the first,
ie, segments A, B, C, ... there are located in the same places as in the first.
Similarly, the fourth disc is an exact copy of the third. Upon receipt of a
request for a device to read data can be read from any of the two disks, so that
each pair of mirror disks on average performs only half the requests. When
recording data written to both disks. If one of the disks fails, access is not
interrupted, and no loss of data occurs. All operations performed by the
remaining operational disk. Since the location of the data is no different from
the traditional, for this configuration there is the problem of "hot spots". A
segment is located on the first and second drives, in the next segment, in the
third and fourth drives. Segment C located on the first and second drive,
followed by A. Segment Segment 0 located on the third and fourth drives next to
the segment B, and so on. Dimensions segment chosen large enough in comparison
to the size of user queries. If a request is received for reading, access only
to those implemented in segments of one of the two disks, which are required
data. If this request for the record, the resort also happens only to those
segments whose address corresponds to a logical address in the request. The
scheme is no longer suffers from the problem of "hot spots", and its performance
markedly higher than previously described configuration. Wisdom of the disks "mirror
pairs can perform queries on reading at the same time. If you select the segment
size smaller than the size of the request, the sequence of requests for large
size can be obtained from the performance close to the productivity RAID-0. The
workload can be balanced between the discs by selecting the size of the segment
in a way that, on average, the best possible performance.
Next part
|