Jawad

Some thoughts on this with reference to the original points you raised.

1/more powerful and robust - emotive argument not factual
2/and a real pain if you need to copy tapes for DR, multiple clients doing
restores at the same (oh their data is on the same tape as someone elses),
does a restore pre-empt any tape copying or do you need to get admins
involved?
3/except the master which must be HA'd somehow. Concept is good though,
loads of incoming backups, one of the backup servers die and all new jobs
get routed to the surviving servers (unless its the master thats gone)
4/you don't get tar if you do point 2. Yes tapes can still be scanned in to
create index entries but talk to anyone who ever ran Legato and they'll tell
you their biggest gripe is the time they spend scanning tapes.

NetBackup has some good stuff - one day it will be horses for courses but
think about automation potential today. Last I heard the incremental forever
'~equivalent' wasn't quite there yet, oh and take a real close look at how
database (index file) backups work. Is it still flip/flop between two named
tapes?   How does this work with DR? Its files not database so you don't get
two-phase commit with whats on tape.I'm about a year out of date so don't
take it all as gospel - check with the reps.

In case there's any confusion - I liked what I saw of Netbackup and believe
that if they solve some core problems it will be a real contender. One area
that it will still beat ADSM is local backup performance with its ability to
really drive tapes to their max. I think some of the developments up and
coming in ADSM will sort that out though - certainly using the ADSM API you
can write code which drives the tapes much faster than the regular baclient
.

Regards
Alan