> If you have large, slow fileservers it is sometimes better to break 
> up the filesystems
> you are backing up into different client savegrps.   Then you can 
> run the backup just
> for some filesystems then run the others at a staggered time or even
> a different day.
> You may create multiple clients with different configs, of the same 
> host and back them up
> in different savegrps.  We did this to cope with backup of a huge 
> slow NDMP job andcit
> works well.

We also set up our job to list specific directories, rather than just use 
saveset ALL. Ours is a Windows file server, but the concept is that by 
listing directories, you get better speeds, since you are doing parallel 
backups rather than serial-style backups. In our case, we listed the dozen 
or so top level folders. 

We also set a (global) directive to skip certain file types (i.e., TMP 
files, LCK files, MDF/NDF files (MS SQL database, which are held open by 
the SQL process, and so won't be backed up anyway), files that start with 
a tilde (~) - on Windows, that indicates either a temp file or an open 
scratch file, media files under the user hierarchy ... things like that.