Having defined the data to be backed up, you should perform backups in accordance with the rate at which you are creating data to be backed up.

The frequency with which you backup should be determined by

• The rate at which you are creating data.
• The cost of losing data.
• The cost of performing a backup.

It's possible to create a mathematical model to formalize all these factors and determine an "optimal" backup strategy. However, such models will not usually reflect the true human cost of losing data. For example, if someone writes a poem or an essay, having to rewrite the poem or essay will usually be stressful to the writer, and the result may not be as special as the original. Similarly, if a potential new customer phones you, and you enter their details into a contact database, and then lose the database, there may be no way of contacting that customer ever again.

A backup failure that requires a whole office to have to redo a week's work can change an entire office's attitude towards computers. If the network can't be trusted, everyone will feel nervous entering any data into their computer and will start regularly printing out everything they do. So regular backups can help save paper too!

All these considerations mean that the best "strategy" is to backup often enough so that if you lose a random disk at a random time, the chances are that you will lose very little data indeed. This usually means daily backups, combined with disk-to-disk copies or mirroring during the day.