Re: [Veritas-bu] Tape Drive Statistics
2009-07-10 12:49:51
On Fri, 10 Jul 2009, Heathe Kyle Yeakley wrote:
> Hello group,
>
> I am trying to performance tune a new library I installed. I have the
> performance tuning guide and I'm looking at my data path. I currently
> have statistics on my tape drive (from the libraries point of view),
> stats from my SAN switch, and stats from my LAN. I'm watching iostat for
> the disk that I'm writing to, but I'd like to see the performance of my
> tape drive from my Master server's point of view. I don't believe iostat
> accepts tape drives. How do you measure tape drives statistics in
> Unix/Linux (specifically RHEL).
>
> On a side note, I'm reading up on how to tune SIZE_DATA_BUFFERS and
> NUMBER_DATA_BUFFERS. If I'm watching statistics in real time for my
> performance tuning backups and restores, can I change these values in
> real time without having to restart Netbackup (or any of my
> master/media/clients) and see the results take affect, or do I have to
> restart services for the new values to take effect? Anyone have tips on
> how to tune these values?
>
> Thank you.
>
> - Heathe Kyle Yeakley
> _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>
There was a paper written by IBM or Symantec awhile ago when they did
testing, the most optimal was 262144 (256k) for the size and 32 for the
number, beyond that (note: this paper is OLD, LTO-2)-- increasing the
number did not buy any improvement in performance.
Don't forget about SIZE_DATA_BUFFER_DISK and _RESTORE to that
NUMBER_DATA_BUFFERS_RESTORE
Also NET_BUFFER_SZ (from clients)-- currently using 1MB here.
Justin.
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|
|
|