How many list subpartitions in each partition ? Are you using ASSM or freelist management on the tables or indexes ? Do any of the indexes have very repetitious key values ? Does the insert include a "sequence.nextval" and if so is that column indexed ? Is the volume of data (including indexes) large relative to the size of the cache ? Regards Jonathan Lewis dombrooks@xxxxxxxxxxx wrote: > 9.2.0.8. > > I have a table partitioned daily and subpartitioned by list on which ther= > e are two local indexes and two global unique indexes. > > > Every day, the first time I run this statement with this new data I will = > get some 3000 "leaf node splits" of which 1020 are "leaf node 90-10 split= > s" then I roll it back. > > Every subsequent time I repeat this statement and roll back, I insert the= > exact same data in the insert .. select but I get 1020 @"leaf node split= > s" and "1020 leaf node 90-10 splits'". > > Any thoughts on the difference? > -- //www.freelists.org/webpage/oracle-l