In previous columns, I've raised the issue that modern disaster recovery strategies require near real-time data replication to ensure that data is protected almost as soon as it is created. Furthermore, the latencies associated with long-haul communications to a disaster recovery site would seem to exclude the use of synchronous mirroring. Under synchronous communications the slowest link in the chain essentially dictates the speed of both the application and the replicated data.
When it comes to disaster recovery the most suitable method is asynchronous replication, where the relationship between the application's data flow and the transmittal of data to the replication site is only loosely coupled. The application continues to use data at its anticipated speed, and data is sent to the remote site "as fast as possible."
Using an appliance to provide IP replication offers numerous advantages. It provides centralized control of disaster recovery processes for all data, regardless of where the data is stored and irrespective of whether the access method is a storage-area network or a network-attached storage. It also enables the use of heterogeneous storage. For example, data on a high-end Fibre Channel array can be replicated to a low-cost ATA array at the disaster recovery site - removing a cost obstacle that has held back some disaster recovery deployments.
An appliance that reads and writes to a heterogeneous storage environment also provides maximum flexibility in designing and deploying protection that is application-specific.
For instance, appliances not only come with software that provides asynchronous (as soon as possible) mirroring, but now also allow point-in-time scheduled replication. Application-specific snapshot agents ensure that the point-in-time replica is a valid "mountable" image of the application data. Deploying a delta-replication strategy (sending only changes since the prior replica) for efficient use of bandwidth, administrators can replicate data frequently to minimize risk.
Better yet, by combining a point-in-time replica of a database with an asynchronous mirror of the transaction data, administrators can rapidly bring the application back online and roll forward to the point of failure with little or no loss in the face of even the worst disaster.
Even midsize companies can use these technologies to design an affordable disaster recovery infrastructure. In fact, virtually any organization with multiple data centers can be "self-protected." Rather than use a third-party disaster recovery site, data can be replicated to low-cost arrays in a peer data center.
Modern business models and current events have conspired to provide greater visibility to every organization's disaster recovery plans. There is tremendous pressure to ensure the safety and rapid recoverability of operations regardless of the threat. With data replication tools such as those that are now available, IT is well equipped to meet the challenge.
Vendors with appliances that can be used in this way
include DataCore (