Should TSM restores for servers running archival backups to be faster or slower than restores from non-logged incremental TDP backups? We’re finding that a restore requiring 6 days of TXN’s takes many hours to complete.
We perform the following backups:
Selective: FRI
Incremental: SAT - THUR
Archival: every 4 hours.
The server in this case has 1500 application db’s and is 3.5TB. The app being restored is 4GB.
Subject: Restoring t-logged databases take longer to restore - if you restore to a point in time
Restoring the full backup should take the same amount of time as a non t-logged database.
The extra time comes from restoring the database to a point of time (like restoring a full Sunday night backup to Friday afternoon.) Each write to the database will be written to a transaction log file. The busier the server, the larger number of transaction logs are created (and backed up.)
When you perform a point-in-time restore each t-log file that contains a write to the database will need to be restored. The log activity is then played back to the restore file. The farther away from the full restore date the more t-logs will need to be restored and replayed. So busier servers will have to look at more t-logs than a less busy server.
We have about 2500 users on 4 servers. In my experience a full restore and replay of 5 days of t-logs takes about 30 minutes.