The following note was written by Kevin Cardew after we had discussed how to reduce the backup time of some of our large databases, especially when they constrained by most of the data being stored in a single bigfile tablespace .
Our build standards require that 11g (184.108.40.206) databases are built using BIGFILE tablespaces, however this has led to one or two problems with RMAN backups. In several of the databases the deployed application, third party supplied, has all the objects deployed to a single tablespace. The result of this is that an increasing amount of the database size resides in a single data file. During the backup RMAN balances the load by attempting to divide the backup volume (data files) over the available channels to maximise backup speed. Given the unbalanced nature of the data files, this results in a single channel being allocated to the largest data file and the rest load balanced across the remaining channels. Therefore the overall backup time is determined by the time taken to backup a single data file. Read the rest of this entry »