We have almost 70 databases spread across over 40 servers, running 9.2.0.8, 10.2.0.4 and 11.2.0.7. We are in the process of upgrading our Oracle Financials databases from 9.2.0.8 to 11.1.0.7. Our servers are a mixture of Sun Solaris 5.8, 5.9, 5.10, Red Hat Linux 4 and 5 and one or two HPUX boxes. We are NFS mounting volumes from one of two storage servers that have 25TB of storage each.

We are running full RMAN backups of the database using a single channel to the NFS mount. Under 9i the best case backup was almost 600GB to the NFS mount in 3 hours 40 minutes. But the average backup is a little over 6 hours. Once we move a database to 11g, the backup can be up to 17 hours. We are not using io slaves and we do have asynchronous io turned on by default.

I am looking for ideas on what I can do to both identify the bottleneck and shorten the amount of time the backup takes. We may at some point look at doin incremental backups, but we should be able to get a full backup in less than 7 hours, especially since we are using compression on the 11g backups which should reduce the amount of io traveling to the NFS mount.

Thanks in advance, Brian