BACKUP POLICIES FOR A DATABASE SYSTEM WITH TEMPORARY LARGE UPDATE

In the batch processing, the updated data in a database system sometimes occur with a large amount in a short time. We propose to make independent backup schedules, which copies the batch-processed data for safety, for the full backup schemes. We apply the cumulative damage process to model database backup schedule, by replacing shock by time for data update and damage by the amount of updated data. When the total rewritten data has exceeded a level K, a full back up is performed. However, the overhead for such a full backup might become sometimes very large in a mass database system. To lessen the overhead for backup actions, we make an independent backup, which copies batch-processed data. The overhead for independent backups increases in proportion to the total amount of updated data. We obtain optimum full backup intervals with independent backups to minimize the expected cost rates analytically