TABLE OF CONTENTS Before discussing the strategies for data recovery and availability it s worth pausing to consider the degree to which this is at all needed. There are essentially two major scenarios. The first scenario concerns itself with the case that a database server fails for any reason including hardware fault or software crash, and while that server is being repaired we require a strategy that allows processing to continue. The second scenario is concerned with possible geographic disaster, where entire neighborhoods or cities can experience a serious disaster that would make the database unusable either because it has been destroyed or because connectivity to the region has been disruptive (a working database that nobody can connect to is less than useful). This would include major disasters such as New York City during the September 11 attacks, or the devastation caused by hurricane Katrina [5] that caused massive destruction.
The major strategies for database availability fall into six categories:
Application controlled multicast.
Replication.
Transaction log in-memory shipping.
Shared disk.
Cluster failover.
Flash copy or backup transfer.
With application-controlled standby server, the application applies all change in the database to two or more database servers. We ve called this application-controlled multicast for want of a better term. As a result at any point in time there is more than one database that has all the required data. Should one database server fail, the remaining servers are available with all of the up-to-date...
