Hello Joost,
What do you mean by "LTO1 media"? We use the atdd driver, ver 3.1.0.3 GES
IBM magstar ultrium tape device driver, the devices show "CLAIMED" on ioscan
to all of the LTO (1 & 2)devices with this software, are you telling me that
with HP-UX you must install two separate drivers and bind each by hand?  

I'm looking at settings by device in the manual David C. pointed me to (many
thanks, David!) I'm investigating whether /opt/OMImag/bin/atdd_cfg numerous
settings might do the trick, until I receive your answer re:diff software.
Our current setting for SIZE_DATA_BUFFERS is currently 262144 by the way.
Thank you again, Aren

-----Original Message-----
From: Joost Mulders [mailto:mail AT j-mulders.demon DOT nl] 
Sent: Friday, December 05, 2003 1:09 PM
To: veritas-bu AT mailman.eng.auburn DOT edu; Aren.Parisi AT wwireless DOT com
Subject: Re: [Veritas-bu] Where in NBU are drive settings (LTO-1 and LTO-2)
tuned separatel y?


Hi Aren, 

Two notes on your question:

1. Media
Make sure you are using LTO2 media in your LT02 drives. With IBM drives (and
I 
think the others as well), using LTO1 media in an LTO2 drives reduces the
max. 
throughput to 20MB/s. 

2. SIZE_DATA_BUFFERS
Make sure you have /usr/openv/netbackup/db/config/SIZE_DATA_BUFFERS file on
your 
media server(s) with a value of 262144 (256K) in it. This file controls the
size 
of the data buffers yadiyadiyadi, but also the blocksize used to transfer
data 
to the drive. Using "big blocks" lowers SCSI/FC bus overhead (each block has
to 
be sent and acknowledged) and it improves the efficency of the compression 
algorithm in the drive: the larger the block, the more "similar" patterns
can be 
found and compressed.

The above is true for all tapedrives. There are no LTO2 specific knobs.

joost


>NBU 3.4.1 Master:HP-UX 11i three-way service guard cluster Library:ADIC 
>scalar10k IBM Ultrium drives:20@LTO-1, 10@LTO-2 same cabinet. We have 
>just benchmarked our new LTO-2 drives and they are NO FASTER than the 
>LTO-1 drives. Are there separate tuneables for the new drives we are 
>not looking at?  Buffer settings, something undocumented perhaps?  We 
>have a 3GB (bound) FC ethernet connection on the media server so I 
>don't think it's a network bottleneck, disk is XP-512 (rebadged 
>Hitachi) screaming fast disk.  Tested same class on each type of drive, 
>identical runtimes. ??? Thank you experts.
>
>     Aren M. Parisi *
> Information Technology
>  Western Wireless Corp.
>(425)313-5225 mob.(206)795-6631
> Where are we going and why am I in this handbasket?
>

-- 
Long may you run.