Raid Implementations - Hitachi AMS 2100 Hardware Manual

RAID implementations

The AMS 2100/2300 supports RAID 0, RAID 1+0, RAID 1, RAID 5, RAID 6,
or an intermix, as described below.
RAID
Level
0
1
1+0
1
5
6
1. RAID 0+1 is described in place of RAID 1+0 in some places in this document.
However, it has the same meaning as RAID 1+0.
1–22
Table 1-10: RAID Implementations
RAID 0 group stripes data across all disk drives in the group to attain higher
throughput. There is no sparing disk drive function with this configuration.
Notes:
• When a failure occurs in a disk drive, data cannot be restored
internally within the disk array because RAID 0 has no data redundancy.
Therefore, although the host receives an error message, the data is left in
its current state.
• Unlike with an independent disk drive, error block management
(assignment of alternative block for the independent disk drive with the
Reassign Blocks command) is not done. Therefore, when a RAID 0 disk
drive becomes inaccessible due to a failure, all the logical units in the
RAID 0 configuration including the failed disk drive become inaccessible.
Restoration of user data with backup data is essential after replacing the
failed disk drive.
• Do not allow RAID 0 to coexist with another RAID group. Do not form
a group with two or more RAID 0 systems, for example (4D + 1P) × 2 +
2D.
• Each time a failure occurs in a RAID 0 disk drive, data that cannot be
written into the disk drive (pinned data) is accumulated in the cache.
When the amount of accumulated data exceeds a certain value, a write I/
O instructed to the other RAID group will not be accepted and the system
will crash. In this case, restoration of user data and resynchronization of a
file system using the backup data are required for the entire array. The
recovery time for this procedure can be quite long.
• Do not adopt the RAID 0 configuration to its fullest extent. Consider
the above notes when building the configuration.
RAID 1+0 groups provide data redundancy like RAID 1 by copying all the
contents of two disk drives to another pair. Different from RAID 1, data
striping is performed for a maximum of 16 sets of two disk drives.
RAID 1 array groups consist of at least two disk drives in a mirrored
configuration. Data is mirrored across the groups of two adjacent drives.
The stripe consists of two data chunks.
RAID 5 uses from 2 to 15 data disks, and has a parity disk performing the
data striping.
RAID 6 uses from 2 to 28 data disks, and has two parity disks performing
the data striping. Configuration with two parity disks provides the
redundancy that can sustain two points failures.
Introduction
Hitachi AMS 2100/2300 Storage System Hardware Guide
Description