Hi,

Everynight, I run ADSM ConnectAgent/EBU to backup a 80GB Oracle7 database
(with EBU parm parallel=4) on machine A to ADSM server disk pool with
"cache migrated file=Y", then restore from ADSM server disk pool to machine
B (also with EBU parm parallel=4).

Everynight (far after the EBU backup completed, sometimes in parallel with
EBU restore), through external scheduler Control-M schdule Unix script to
migrate disk pool data to 3590 tape (collocate=y). The script basically
1. "upd stg hi=0 lo=0 migprocess=1", to start migration
2. wake up every 5 minutes to monitor until migration completed,
3. "upd stg hi=90 lo=80" to stop migration

One night, the script failed due to migration process failed and step 3 was
not executed, EBU backup on the subsequent night go to disk then migrated
immediately; that's reasonable. Unexpectedly, the EBU restore from ADSM
tapes. As only one ADSM node was used, at every instant, only one EBU
streams got the tape leaving the other 3 in media wait. Even it got the
tape there spent some time to locate file on tape as the file sequence on
ADSM tape is probably not the same as EBU file restore sequence. The
restore prolonged from 4 hour to 9 hour.

As ADSM keep the most recent files on cache, so normally the restore
involved no tape. But for that odd night, files immediately migrated were
not remain in cache.  Question:

1.should the most recent files be cached irrespective when it is migrated?
2. there are backup on several large Oracle8 databases (900GB, 450GB,
230GB...etc), although no regular restore. Would like to share experience
on speeding up EBU and or RMAN restore from ADSM tapes.

Thank you in advance.

Regards,
Eric Tang