Managing Database Backup and Recovery - High-Performance Web Databases: Design, Development, and Deployment - page 691

system still functions by using the redundant data found on the surviving
drives. There are two types of RAID 5 available today, namely, software- and
hardware-based RAID 5. Hardware RAID is the more desirable implementa-
tion method because it was designed with drive failures in mind. Extending
the tolerance level of a RAID 5 system can then be achieved by mirroring or
duplexing drive arrays. This type of extension allows for whole drive arrays
to fail without impacting the system

Duplexing

Duplexing is similar to mirroring except that in a duplexed configuration
separate controller cards manage each drive or sets of drives. In essence,
duplexing is Raid 0 with an additional layer or redundancy. The second disk
controller cards remove a single point of failure that is exhibited in a stan-
dard mirroring (Raid 0) configuration.

Partitioning

Partitioning is the ability to deploy a database system across multiple
servers where each server houses a different portion of the overall data-
base. Should a server go down, only the component running on that server
becomes unavailable. In this scenario the database can continue to func-
tion normally, provided applications are written to handle these types of
situations. Additional protection can be achieved by employing RAID 0,
RAID 5, or duplexing to minimize system downtime further.

Replication

Replication offers the ability to publish the contents (complete or por-
tions thereof) of a database to another or multiple servers in an environ-
ment. The technique is similar to partitioning; however, to employ
replication requires sophisticated application transaction logic to be used
effectively. Replication allows for the mirroring of database transactions to
be replicated in a secondary database at the central site or in a distributed
location. Ideally, all transactions should be processed at a central database
and the transactions should be replaced to the other subscribing sites. This
eliminates the difficulty that becomes inherent with transaction logic of the
traditional two-phase commit that fails as a result of hardware failures.

Clustering

servers to share or cooperate
with each other in utilizing common resources. Clustering allows systems
to monitor each other and, in the advent of failure, transfer processing to
their counterpart. Clustering is a very reliable method for maintaining a
fault tolerant and highly available systems environment; however, vendors
approach clustering differently. It is recommended that organizations

Clustering is the ability of a group of n