April, statspack has problems with large numbers, that is a well known =20 fact.Try running it every 10 minutes, on the level 9. You will no =20 longer have problems with the numbers. You will have problems with =20 STATSPACK itself. Second, try putting /*+ APPEND */ into your inserts, I believe it might reduce the number of undo block written. Third, create tables as =20 "nologging", which works in conjunction with the direct load ("append" =20 hint) On 11/15/2004 09:27:59 AM, aj wells wrote: > Okay... more creativity... >=20 > I have a multi TB database (9.2.0.4) that is pulling statspack stats > every hour. > For the first hundred hours of the massive load into these tables, =20 > the > statspack report shows HUGE undo bytes written (which I would expect) > excpet they are negative numbers. negative 3 billion, negative 2 > billion... with trans table gets along the same lines (negative > millions), wraps, shrinks and extends are all negative.... even when > the undo bytes written are positive, the others are all negative. >=20 > Huh? --=20 Mladen Gogala Oracle DBA -- //www.freelists.org/webpage/oracle-l