Replication Mode

The selection between synchronous or asynchronous mode is made according to the RPO/RTO ratio (described in the following paragraph), the distance and the level of availability desired. The quality of link between the sites should also be considered.

The replication technologies can be applied at the storage level; replication is performed between two remote disk arrays connected through Fiber Channel (FC) technology, but it is also possible to perform a data replication through conventional TCP/IP network.

• Mirroring or Cloning

In synchonous mode, this technology provides an identical copy of the production volume, it is generally used for disaster recovery on another site, for tests or backup on a remote server executed from production data.

Clustering

This technology aims to increase the availability of applications which are critical and compatible with this one. The principle is to immediately have a second server in case of failure of the first one. There is an automatic switching over of the application to a redundant or standy server (Failover) through a process of mutual verification of proper operation if any component of the software or hardware supported and supervised by the clustering software were to fail. This mechanism allows to reduce or to make completely invisible to users a major failure of one of the systems. The level of availability is thus significantly increase. Once the component is replaced and the server is back in production, the restoration of a failed system component's share of a load to a replacement component after a failback event (failback) can be executed at administrator's request. We distinguish clustering solutions in "active-passive" or "active-active" mode. In the first case, one of the servers (or node) is standby and ready to support the application executed on the active node. In the other mode, both nodes or n -nodes are simultaneously active and share the application workloads. The level of availability is highter than an active-passive solution, however the cost and complexity of such a solution are much higher. It is also possible to implement extended clustering solution (stretched cluster). In this case, the cluster nodes are on geographically remote sites. Such a complex and onerous solution is for applications requiring very high availability of service.

Virtualization

The concept of virtualization is to abstract physical hardware layer from the point of view of the operating system to optimize the use of resources and to facilitate administrative tasks.

This technology can be applied to different components of the architecture. First, Storage Virtualization by FalconStor provides a flexible approach to data management – move, store, protect, and recover data – without tying your business to specific hardware, network or interconnect protocols. The same virtualisation can also apply to the Ethernet network, workstations and servers finally ( VMware, Microsoft Hyper-V, Xen). The main applications of server virtualization are:

• the consolidation enables the limitation of the number of physical machines which should be managed

• the test and the development by fastly providing ready-made servers upon request or which are exact copies of the production servers at a given time

• business continuation in case of a sinister by supplying normal servers which are immediately operational, this reduces the cost of purchasing spare equipment (spare) necessarily identical to the production server.

Restoration

Restoration is the last chance in case of partial or total computer data loss, it should therefore be flawless. The need for restoration can be expressed in according to 3 distinct needs:

• restoration of the full image of a server (Bare Metal Restore)

• to restore a file previously backed-up or archived which was lost or altered

• migrate backed-up or archived data of a system to another system

To meet each of these needs, it is essential to define two precise objectives, namely Recovery Time Objective (RTO) and Recovery Point Objective (RPO).