On Dec 4, 2008, at 18:53 , Zoltan Forray/AC/VCU wrote:

Are you mis-reading/writing that info?    350GB takes 1-hour to
expire and
about the same to backup??

My 190GB DB used to take 48+ hours to run expire (350M objects) and
2-hours for a full backups!  With recent cleanup/load balancing by
moving
nodes to a new server, the expires are down to 24-hours.

Yeah I find that pretty impressive, since I think it is quite
difficult to optimize the DB disk subsystem to have both fast backups
(sequential I/O, read-ahead/cache on) and fast expiring (random I/O,
read cache off). I'm even thinking of a hack to automatically
temporarely enable read caching in DS SAN during backups. Expiring is
propably somewhat CPU bound too, maybe fast Windows boxes are good in
that sense.

Jeremy, have you tried to restore the DB to your RAID5 array? Doesn't
happen in an hour, doesn't it? When it comes to DB backups, who cares
if backups take 1 or 3 hours, but restore speed is everything..