Using nvSRAM in RAID Controller Applications
as airline reservation systems, financial and banking
applications, etc.
RAID Level 6
RAID 6 is a non-Berkeley level controller that is
designed for extremely high data reliability. RAID 6
is an independent access array concept that
requires two parity blocks be updated for each block
written. This requires an extra parity disk but gives
nvSRAM Applications & System Archi-
tecture
In modern RAID systems the Array Management
Software can run in the host or in a dedicated
embedded controller. Most modern systems are
using embedded controllers, including many manu-
facturers using the Intel i960 chip as the engine.
the added data safety of requiring 3 disks to fail
before data will be lost. RAID 6 data transfer and I/O
capability is lower than RAID 5 for writes, but data
reliability is highest of all RAID architectures. Pres-
ently RAID level 6 is not widely used because of the
higher costs associated with the added complexity,
and the high penalty paid in system I/O performance
due to long write times.
In the past RAID systems were designed to use a
distributed block of disk to maintain system configu-
ration and to store system recovery address vec-
tors. The primary problem with this type of
architecture is that if a power failure occurs, and the
controller’s volatile system memory is lost, the entire
disk array must be scanned upon power up to rees-
tablish configuration and to redefine data locations.
On a large array this is very time consuming, requir-
ing many minutes to accomplish. Service-oriented
industries cannot afford this length of down time and
must come up and be operating very quickly once
power is restored. In the latest generation of RAID
systems the restart vectors are stored in nonvolatile
semiconductor memory on the controller board
itself. Due to the fact that the Array Management
System is constantly moving data among the indi-
vidual array members to optimize I/O balance, maxi-
mize I/O rates, and assure redundancy, the RAID
controller is constantly tweaking the address vector
tables. Also, the system configuration data is being
Additional RAID Implementations
RAID 10 is a combination of RAID 0 & 1. This archi-
tecture gives high I/O performance and good data
reliability. It is accomplished by using RAID 0 (data
striping) to enhance I/O rates and by using RAID 1
(disk mirroring) for high data reliability. RAID 10
requires costly hardware (disk and port) to imple-
ment, and is primarily used in applications where the
data has high value and can justify a mirrored stor-
age system.
RAID 53 is a combination of RAID levels 0 & 3 and
provides RAID 3-like data transfer performance, and
striping-like I/O request rates at RAID 3 or 5 costs.
RAID 53 is used where both high data request rates
and high data transfer performance is required such
Physical
Disk 0
Physical
Disk 1
Virtual
Disk
Chunk 0
Chunk 4
Chunk 8
Chunk 12
Chunk 1
Chunk 5
Chunk 9
P (12-15)
Physical
Disk 2
Chunk 0
Chunk 1
Chunk 2
Chunk 3
Chunk 4
Chunk 2
Chunk 6
P (8-11)
Chunk 13
Chunk 5
Array
Management
Software
Chunk 6
Physical
Disk 3
Chunk 7
Chunk 8
Chunk 9
Chunk 10
Chunk 11
Physical
Disk 4
Chunk 3
P (4-7)
P (0-3)
Chunk 7
Chunk 11
Chunk 10
Chunk 14
Chunk 15
Figure 2
Example of a Typical RAID Level 5 Controller
From The RAID Book Edition 1-1
8-27