Veritas-bu

[Veritas-bu] [NBU 3.4.1] Multiplexing

2003-12-12 10:22:54
Subject: [Veritas-bu] [NBU 3.4.1] Multiplexing
From: larry.kingery AT veritas DOT com (Larry Kingery)
Date: Fri, 12 Dec 2003 10:22:54 -0500 (EST)
> "The per Class settings for max. multiplexing work only for that
> class, so if more than one class runs at once it will multiplex up
> to the STUs max."
> 
> No, not true.  It will only MPX up to the lower of the two.

Actually, it's more complex than that.  It could mpx up to the higher
of the two schedule settings, but no more than the stu setting.  It
all depends on which order they get started (if the mpx=2 start first,
the mpx=6 jobs may join them; if the mpx=6 jobs start first, the
mpx=2 will not).  The schedule setting says, basically, "I will not
enter a mpx group which ALREADY has N active streams" -- it doesn't
say anything about limiting OTHER jobs which start later.

To stop this, the simple answer would be to use different pools.  I'd
probably start by determining whether the difference in mpx levels is
going to have any (negative?) effect on restore in the first place.


> 
> For details on this see the Chapter in the NetBackup SAG titled
> "Additional Configuration".  There is a section called "How to
> Configure Multiplexing" that explains this and gives a good detailed
> example.
> 
> 
> - Scott
> 
> 
> 
> 
> 
> Arne Kloecker <arne AT topnet DOT de>
> Sent by: veritas-bu-admin AT mailman.eng.auburn DOT edu
> 12/05/2003 05:31 AM
> 
>  
>         To:     veritas-bu AT mailman.eng.auburn DOT edu
>         cc: 
>         Subject:        [Veritas-bu] [NBU 3.4.1] Multiplexing
> 
> 
> Hi,
> 
> to reduce restore time for some special classes, no more than 2
> streams should be multiplexed per drive. Other Classes should be
> able to multiplex more streams.
> 
> The per Class settings for max. multiplexing work only for that
> class, so if more than one class runs at once it will multiplex up
> to the STUs max.
> 
> Can i have some classes which write only 2 streams per tape and
> others up to 6 (or whatever)?
> 
> Any idea is apreciated.
> 
> Thanks in advance.
> 
> Arne
> 

-- 
Larry Kingery 
                     Of couse, I could be wrong.

<Prev in Thread] Current Thread [Next in Thread>