Overhead of recovery

Although the value of recovery is obvious, it is not without cost. Logging, for example, takes time and space: some amount of processing time is required when a log record is created, and each log record uses space on a data set. Each checkpoint written also takes time and space, as does each backup copy made.

Having a basic recovery scheme is a practical necessity, not really a matter of choice. However, you do have a choice about how elaborate to make your recovery scheme.

Generally, the more extensive recovery preparations you build into your daily operations, the faster you can bring your system back into production after an error occurs. The trade-off is between speed of recovery and the daily overhead of recovery maintenance. You also need to consider the cost of having your online system down.

In judging this trade-off, consider how frequently errors occur that require recovery. Errors will certainly occur, but rare errors merit less recovery preparation than those that occur more frequently. You decision about the extent to which you prepare for recovery mechanisms should be based on knowing your installation's needs and priorities, and weighing them against each other.