ICP GDT RS Series, Руководство пользователя

Страница: 66 / 186

66
C.6.2 Which Level of Redundancy is Needed ?
To come straight to the point, RAID 0 (data striping) does not imply any redundancy at all
(the R in front of the AID is rather misleading). On the other hand, a RAID 0 disk array is
pretty fast, since no parity information is required. With RAID 1 (disk mirroring), the data is
100% redundant because it is mirrored. This is definitely the highest level of redundancy,
but the most expensive one, too. An interesting combination of RAID levels 0 and 1 is
RAID
10. Two RAID 0 stripe sets are simply mirrored. If one drive fails, the data are still available
on the mirrored drive. With RAID 4 (data striping with dedicated drive) and RAID 5 (data
striping with striped parity), parity information is calculated from the present data with a
simple mathematical operation (eXclusive OR, XOR), and stored either to one dedicated
drive (RAID 4) or to all drives (RAID 5). If one drive should fail, the data of the defective
drive can be reconstructed on the basis of the normal user data and the previously calcu-
lated parity data. RAID levels 4, 5 and 10 can tolerate the failure of one drive just as RAID 1,
but in comparison to the latter, RAID 4, RAID 5 or RAID 10 are less expensive.
As already mentioned before, the entire disk array controlling function is carried out at con-
troller level and therefore does not load the host computer.
Let us have a look at the following table which explains the correlation between the RAID
level, usable disk capacity and number of physical hard disks. To make things easier, we
consider identical 1 GB hard disks:
Usable storage capacity of the disk array
RAID Level 2 hard disks 3 hard disks 4 hard disks 5 hard disks
RAID 0 2GB 3GB 4GB 5GB
RAID 1 1GB 1GB 1GB 1GB
RAID 4 - 2GB 3GB 4GB
RAID 5 - 2GB 3GB 4GB
RAID 10 - - 2GB -
It is quite obvious that the redundancy of level RAID 1 soon becomes very expensive when
more than 2 hard disks are used. Only with RAID 4 and RAID 5 have you a reasonable rela-
tion between storage capacity and expenses for the disk array.
C.6.3 Do we Need Hot Fix drives ?
In other words: Should RAIDYNE® automatically reconstruct the lost data after a hard disk
failure ? One of the reasons that have led you to choose
R AID disk arrays definitely lies
with the r edundancy, that is, the data security you still preserve even in the event of disk
failure, thus resting assured against loss of data and time. Hot Fix drives are possible with
all RAID 1, 4, 5 and 10 disk arrays. In order to assist the following considerations, we define
the term
time without redundancy , TWR . Set apart the time needed to set up the disk array
(state build ), the time without redundancy should be kept as short as possible. Let us as-
sume that one of the hard disks of the RAID 5 disk array we set up with example 1 fails. The
disk array is without redundancy. TWR starts to run. Any superfluous prolongation of the
TWR (because you have to get a replacement drive, or because you did not realize the fail-
ure immediately since you didn't hear the ICP Controller's alarm signal, or because nobody
checked the file server) increases the risk of data loss which will occur if a second drive
should fail. Therefore, new redundancy should be created as soon as possible and in an
entirely automated manner. Integrating a Hot Fix drive as an immediately available and
auto-replacing drive is the only way to keep the TWR as short as possible. Only a Hot Fix
drive can ensure optimal disk array security and constant data availability. Of course a Hot
Fix drive is not compulsory. If you control the disk array at regular intervals and immedi-
ately replace a defective drive (by shutting down the system or hot-plugging), you can do
without a Hot Fix drive.