[Veritas-bu] Using MPX causing disk access times to skyrocket
2002-11-14 13:23:34
Subject: |
[Veritas-bu] Using MPX causing disk access times to skyrocket |
From: |
Geibel, Jonathan" <Jonathan.Geibel AT disney DOT com (Geibel, Jonathan) |
Date: |
Thu, 14 Nov 2002 10:23:34 -0800 (PST) |
so that's disk *access* time not necessarily disk throughput.. one guess
is that when you multistream from one disk to two drives (each drive
accepting two streams) netbackup is going to have to bounce between
writing stream 1 and writing stream 2.. it's not a very intelligent
streaming mechanism if I'm remembering correctly..
because it's just load balancing time between the two streams, the disk is
having to constantly switch the location where it's reading it's data --
therefore increasing the overall access time.
when you do one stream per tape drive it simply streams the data straight
through until it's done. as soon as you do multiplexing you start causing
this type of behavior where the disk head is going to have to bounce
around between two areas which will increase those vxstat statistics..
just an idea,
Jon
On Thu, 14 Nov 2002, Mark Evangelista wrote:
> One plausible explanation is that multithreading is keeping your tape speeds
> up which will improve throughput.
>
> "W. Curtis Preston" wrote:
>
> > Instead of sending 10 streams to 10 tape drives, we're sending 10 Oracle
> > streams
> > to five tape drives.
> >
> > The DBAs are claiming that vxstat shows that disk access times have shot
> > up significanly during backups, and this doesn't make any sense to me.
> > Does anybody have any idea why this would be the case?
> >
> > _______________________________________________
> > 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
>
|
|
|