Restore-Backup process for Self-Driven database

[alex-dukhno]

Backup prediction in self-driven database

• by observing load on database engine predict "best" time to trigger backup process (a fancy way to describe a linear regression)
• by observing data changes pattern in table(s) (append-only inserts, range delete's or update's) choose backup tactic:
  backing up full table
  backing table page by page
• by observing load pattern on database automatically change strategy
  full database backup
  incremental database backup

pseudo real example for testing or demonstrating purpses:

• take existing (tpc-c) or create your own benchmark that has append-only insert, update's delete's queries (maybe by index range), read-only tables (maybe after "some point" in benchmark)
• run benchmark to trigger backups
• in the end run new instance of database and restore data from backup
• compare content of databases