Exactly, I defer to one of my original replies to this thread.

"WORD TO THE WISE:  Do the math before you set your MPX value...What is your 
network speed, what is your drive speed, how many MPX before you start seeing 
diminishing returns.  If you are sending too many MPX streams to the drive you 
will end up with WAITING FOR EMPTY BUFFERS in your bptm log on your media 
server, thus meaning that your drive can't keep up.  Alternatively speaking if 
you have WAITING FOR FULL BUFFERS, then perhaps you can increase your MPX 
value, but remember balance is the key when you look to tune your environment."

Thanks Larry.

Chapa

Quoting Larry Kingery <larry.kingery AT veritas DOT com>:

> Even benchmarks won't help in most cases.  There's just too many
> variables involved, YMMV and all that.  You just can't say whether MPX
> will make restores faster, slower, or have no effect.
> 
> I still wish I knew if you had to keep DLT drives streaming DURING
> READ.  This is a clue to one way MPX might actually be faster.
> 
> Imagine a client with two disks, each with it's own filesystem and
> able to read/write at 5MB/s.  The network is able to transfer data at
> 10MB/s.  There is one tape drive on the master/media server capable of
> 10MB/s.  Now, which restores faster in this situation, MPX=1 or MPX=2
> (where the two filesystems are backed up at the same time)?
> 
> david AT dsihost-srv01 DOT com writes:
> > John you are absolutely correct!  I have no benchmark data but I don't
> think 
> > you really need it to argue that a restore of a single contiguous stream
> will 
> > be faster than that of its peer that has been Multiplexed with 8 other
> jobs.  
> > 
> > However, I believe that is something that needs to be decided at the
> Business 
> > level with regards to each individual site's SLA they have with their
> customers 
> > (user community).
> > 
> > David
> > 
> > 
> > Quoting John Meyers <john.meyers AT wright DOT edu>:
> > 
> > > 
> > >  Regarding performance issues with multiplexing, I would think that the
> > >  higher the number of streams multiplexed to a tape, the slower your 
> > >  restore times would be due to the extra tape positioning required.
> > > 
> > >  I was just wondering if anyone had any data in this area regarding
> > >  performance metrics associated with multiplexing multiple streams to
> > >  tape?
> > > 
> > >  Thanks,
> > > 
> > >  John
> > > 
> > > --
> > >  John Meyers
> > >  Computing Services
> > >  Wright State University
> > >  E-mail: john.meyers AT wright DOT edu
> > > _______________________________________________
> > > Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> > > http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
> > > 
> > 
> > 
> > 
> > _______________________________________________
> > Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> > http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
> 
> -- 
> Larry Kingery 
>              Press <CTRL>-<ALT>-<DEL> to continue ...
>