Redundancy

The traditional term of redundancy is to create an architecture called No Single Point of Failure(NSPoF). The availability of the system is guaranteed by the fact that there is no single breakdown which can cause a service unavailability. Thus, a redundancy system to support a simultaneous breakdown of the N component does not work at N+1. Therefore, this system cannot be defined as highly available as there is a subtle or permanent service unavailability.

The objective of defining a highly available information system is to allow users to work beyond the usual usage frequency of the information system. It is essential to define preventive and protection measures to limit the downtime and to implement rapid methods of restoration.

Replication

The aim of online data replication is not only to guarantee a second copy of data identical to that of production, but to guarantee the possible reconstruction and exploitation of copies which have survived the disaster. Files system or database replication tools should reflect the integrity of the data replicated so that these can be used in case of disaster. These tools should integrate mechanisms called Write Ordering or of Crash Consistency. Replication technologies can be classified in two categories:

• Point-In-Time Copy (Snapshot)
• Real Time Copy

Point-In-Time-Copy

This technology is commonly known as Snapshot and is often associated with a backup application, if possible to a remote site from where the data have been replicated. From the moment the snapshot has been created, original data are copied into the snapshot volume before being modified on the production volume. The snapshot volume is progressively filled as well as writing operations are generated on the production volume until it is filled or destroyed. The application reads and writes seamlessly on the volume of production, the backup application has a read-only acces to the volume of data containing the original data. It may be possible to keep snapshot online to accelerate the restoration process which can be slow depending on the relevance of data to restore.

Real Time Copy