> -----Original Message-----
> From: Thomas R Rhodes/TRR/CC01/INEEL/US [SMTP:TRR AT inel DOT gov]
> Sent: Thursday, 10 May 2001 3:49 AM
> To:   Veritas-bu AT mailman.eng.auburn DOT edu
> Subject:      RE: [Veritas-bu] NBU Data Buffer sizes
> 
> I have been following this thread and have some questions.
> 
> If  you are being delayed should you then look at adding more buffers
> 
        [Marelas, Peter]  Yes. Tuning block size is also important.

> and if you are waiting for full buffers should you then look at smaller
> buffers?
> 
        [Marelas, Peter]  Yes, if its consistent and you need the memory
elsewhere.

> Is waiting for full buffers a problem other than you might not be
> streaming
> your tape drive.  This just means that you tape drive is keeping up with
> the network  or disk.   However, if you are being delayed then the tape
> drives are the slow part of the backup.
> 
        [Marelas, Peter]  Right.

> Am I on track here?  If so it seems I should worry a lot more about being
> delayed than waiting for a full buffer as far as number and size of
> buffers.
> 
        [Marelas, Peter]  Right. Being delayed because there are no empty
buffers
        is what slows down backups when everything else is keeping up.

> Also in looking at my  bptm log I am seeing about 6000 kb/sec on
> completion
> of a backup or a duplication.  I am running STC 9840 tape drives is this
> about what I should expect?
> 
        [Marelas, Peter]  Depends. Is it fibre attached or SCSI? Is it
fabric or loop?
        How many devices on the loop? How many devices on the SCSI bus?
        Is your disk reading faster than 6MB/sec? Is your system lacking
CPU/memory?
        Have you tested the theoretical throughput to the 9840 head given
your
        hardware/environment? Is this throughput greater than 6MB/sec?

        Just questions in no order, that should give you an idea where the
problem
        lies.

        First thing I would look at is Netbackup buffer settings as the
defaults I dont believe
        are optimal for high speed drives.

        Regards
        Peter Marelas