Admins,
I have recently implemented an additional backup routine on our
clusters ( 3 node Tru64 5.1B Pk3 with MC interconnect).
Two nodes have 4 x 72Gb disks mounted on in the local drive cage.
These disks are concatenated into one large AdvFS volume using addvol.
The third node is used to drive the backup using a script which
first uses a modulo of the day number ( 0 = Sun etc ) to decide
whether to mount the backup_domain 0 or 1.
Once mounted, a disk to disk backup is performed.
(The script is listed below)
This backup is taking longer than the "standard" backup to SDLT
tape. The tape backup is taking ~4 hours and the disk backup
~5.5
My question is, can this backup performance be improved by increaseing size and number of memory buffers ( -F and -b switches), or is this simply a result of data being passed through the memory channel interconnect?
TIA
Lawrie.
Received on Mon Nov 10 2003 - 09:23:43 NZDT