On Mar 29, 2007, at 5:41 AM, Ibán Bernaldo de Quirós Márquez wrote:
We are running 4 dscm sel -filelist in parallel to reduce
processing time and increase backup time in a node.
The problem is that 3 process are running at a very good speed but
the last one is running very slow, we have check network
performance and seems to be good. Also we have follow the
performance tuning guide to set network parameters on client and
server...
Well, we can't offer any evaluation without details...such as whether
there is a disk storage pool receiving the data or number of tape
drives available to the client (MAXNUMMP). It is likely that your
four simultaneous sessions are excessive relative to the server
resources that can be given to that client: check your Activity Log,
or perform Query SESSion as they are running, or process the TSM
accounting records to report where wait times are. If the backups
are going directly to tape, then you have to wait for drives to be
available and tapes to mount - and you may be overly consuming
scratch tapes where TSM will see that the latest Filling tape for
that client/filespace is in use, and will use scratches for
additional sessions.
Richard Sims