backup exec 15 on lto6 performance

[bronzemaxell]

i am using backup exec 15 on windows 2012 r2 server,

the storage is overlandstorage XSR120, connected directly to neoseries 8 slot loader with one lto6 drive by SAS cable.

the backup exec 15 talks to the tape drive using linux remote media agent that is installed on XSR120.

the data is roughly 2TB, mostly small sized files  and took roughly about 3.5 hours to save the data to lto6 cartridge, at the job rate of about 10,000 MB/minute.

i am looking to see if there is any room for improvement on backup speed, with the goal of reduce backup time, and identify the bottleneck.

on XSR120 using iostat -m 1 command and disk activity flucutate between 180 to 220 MB/s during backup, the paper said XSR120 can go up to 350 MB/s

the only thing i see on backup exec 15 tape driver property  that can be change is the

block size 64k,

buffer size, i changed to 1MB,

buffer count: 10,

high water count: 1,

read single block mode: disabled,

write single block mode: disabled,

read scsi  pass-through mode: enabled

write scsi pass-through mode: enabled

[Gurvinder]

You can try to change the block size according to this article and see how it goes
https://www.veritas.com/support/en_US/article.TECH69559

[Larry_Fine]

I believe that the standard answer is to use the largest block size and bufer size that are supported in your environment(s) and leave the other setttings at the default.  Changing anything to enhance performance may comprimise readability or restorability.

The concern with the larger block & buffer sizes (which definitly increase performance) is that your disaster recovery sites & scenarios also need to support the same block and buffer sizes or else you could end up with unreadable tapes.  The max block and buffer size supported is impacted by your SAS and FC HBAs (and their drivers), which may or may not be the same at all sites & servers.