Hi, 

Fragment size is a bit of an ancient option that has no advantage for modern 
tape drives like [s]dlt, lto and stk's 9000 series.

Fragment size is available to do 'fast' restores for single files with 4mm DAT 
or 8mm Exabyte equipment and the like. These tape drives do not know about the 
"locate block" scsi command. When NetBackup needs data for restore from these 
drives AND the backup is fragmented, NetBackup can say "fast skip forward n 
fragments" to position the tape as fast as possible near the file.

For newer equipment, the offset of blocks is stored in an index on tape (dlt) 
or 
in a memory chip on the cartridge (lto). NetBackup also 'knows' (by catalog) in 
which block a certain file lives. In this case, NetBackup can say "locate block 
n" and the drive knows by the index the offset of the block and can wind to 
that 
block very fast. So, even if you would have fragments here, NetBackup will 
still 
use "locate block" to do the tape positioning, so having fragments do not 
improve anything here.

On the other hand, fragment size is valid disk stu's which live on file system 
with a limited maximum file size. 

See also,  
        http://seer.support.veritas.com/docs/239409.htm
        
Best regards, Joost


>What (if anything) should you set the max fragment size on UNIX and WINTEL
>when multiplexing your backups. I want to make sure that I don't harm the
>backup performance by playing with the fragment size.
> 
>  _____  
>
>From: veritas-bu-admin AT mailman.eng.auburn DOT edu
>[mailto:veritas-bu-admin AT mailman.eng.auburn DOT edu] On Behalf Of Coco, 
>Samuel
>(Cont, ARL/CISD)
>Sent: Tuesday, February 17, 2004 11:20 AM
>To: Donaldson, Mark; Dave Markham; veritas-bu AT mailman.eng.auburn DOT edu
>Subject: RE: [Veritas-bu] Max Fragment size
> 
>We have been using a 2 Gig Fragment for about 7 to 9 months now and have
>seen very good results.
> 
> 
>Thank you,
> 
> 
>Samuel J. Coco, STG
>Functional Area Manager, ARL
>Sr UNIX Administrator
>BELL      301 394-1151 DSN 290
>CELL (B) 240 398-7121 / (H)443 496-1623
>HOME     410 604-2415
>Email  scoco AT arl.army DOT mil <mailto:scoco AT arl.army DOT mil> 
>-----Original Message-----
>From: Donaldson, Mark [mailto:Mark.Donaldson AT experianems DOT com] 
>Sent: Tuesday, February 17, 2004 11:54 AM
>To: 'Dave Markham'; veritas-bu AT mailman.eng.auburn DOT edu
>Subject: RE: [Veritas-bu] Max Fragment size
>Each fragment is separately tracked in the catalog so having chunks too
>small makes for larger catalogs.  I think there's some header information
>on-tape per fragment, too, so backup sizes might increase a bit.
> 
>Still, I think fragments are a big improvement to restore times are are well
>worth it.  2 to 4 Gig seems to be a common suggested size.
> 
>-M
>-----Original Message-----
>From: Dave Markham [mailto:dave.markham AT icl DOT net]
>Sent: Tuesday, February 17, 2004 4:50 AM
>To: veritas-bu AT mailman.eng.auburn DOT edu
>Subject: [Veritas-bu] Max Fragment size
>I have seen in a few docs about setting this max fragment size to say 2 gig
>or anything less than the max file size your OS can handle. How I understand
>it is that this will then improve restore times as it can skip along these 2
>gig chunks until it finds the section it needs and then begin a read through
>of that chunk to find the files needed and start restore.
> 
>My question is 1. Am I right in my thinking? and 2. What would be the
>downside to this? If it improves restore times what reason is there to not
>set it?
> 
>Thanks
> 
>Environments:-
> 
>Netbackup 3.2, Solaris 8 Using SUN L1000 4 DLT 7000 tape units
> 
>Netbackup 4.5 FP6 Solaris 8 Using SUN L25 and LTO drive
> 
>Dave

-- 
Long may you run.