Hardware Management - Deploying and Managing a Cloud Infrastructure - page 239

performance of write operations can be degraded because data must be written on multiple
drives. Moreover, the slowest drive determines the write speed. As long as at least one of
the drives from the array is functioning, the array continues to function.
RAID-2 Level 2 comprises bit-level striping using dedicated Hamming-code parity. Data
is striped and disk spindle rotation is synchronized such that each sequential bit is on a dif-
ferent drive. This technique enables write and read parallelism by striping input and output
data, respectively. Hamming-code parity is stored on a dedicated parity drive. RAID-2 is
not used in commercial systems, but it sets a precedence for RAID-3 and onward.
RAID-3 Level 3 uses byte-level striping with dedicated parity. As with RAID-2, disk
spindle rotation is synchronized and the data is striped to place each sequential byte on a
different drive. Parity is calculated in bytes and stored on a dedicated parity drive. RAID 3
is not commonly employed.
RAID-4 Level-4 uses block-level striping with dedicated parity. RAID-4 has been replaced
with a proprietary implementation called RAID-DP. RAID-DP stands for RAID-Double
Parity and uses two parity disks (redundancy of parity).
RAID-5 This level uses block-level striping with distributed parity. Parity information
is distributed among drives to minimize the loss of all parity information in case the par-
ity disk fails. RAID-5 requires that all drives but one must be present to operate. In case
a single drive fails, subsequent read requests can be served by calculating from distributed
parity. This way no data is lost, but the read performance is substantially degraded until
the failed drive is replaced. RAID-5 requires at least three disk drives. However, it comes
with some complications: it requires long array rebuild times, and there is a higher chance
of failure during rebuild.
RAID-6 Level 6 uses block-level striping with double distributed parity. An upgrade from
the distributed parity of RAID-5, double parity provides fault tolerance of up to two failed
drives. This enables the creation of larger RAID arrays, especially for systems with high-
availability requirements. If drives are used from multiple different manufacturers, it is pos-
sible to minimize chances of problems associated with RAID-5. With larger drive capacities
and larger array size, it is important to choose RAID-6.
RAID systems are suitable for use with tier 0 and 1 systems because it adds more flex-
ibility and redundancy. In fact, most tier 0 and 1 systems already utilize one or more of the
RAID levels in various storage clusters.
File Systems
File systems are a critical part of the overall system. They are used to manage stored
data and control how it is retrieved. A file system maintains segregation of different files
and manages information about where different fragments of the same file are placed.
Sophisticated file systems that offer high availability, reliability, redundancy (in an event
of failure), and capacity are even more relevant for cloud computing applications.