Some of these loads take 15 hours or more to process. It is a lot of data with at least one CLOB field for one of the tables. There are times that a record or records can be inserted into three different tables. If the process fails after 14.5 hours have lapsed and I have to restart at the very start of the process than that is 14.5 hours that I have wasted. I am estimating that it is going to take me close to 3 to 4 weeks to load the entire set of data. That means I cannot waste any of the time that I have. After a subset of data is loaded then I have to sync 5 intermedia indexes with two of those indexes being User Datastore indexes. One of these indexes can take at least two days to sync for one subset of data. It is indexing fulltext documents that are stored outside of the database as well as the metadata that is stored in the database. I can decrease the subset of data but I want to exhaust all of my other options first. Sandra _____ From: Jared Still [mailto:jkstill@xxxxxxxxx] Sent: Tuesday, August 16, 2005 11:57 AM To: Arnold, Sandra Cc: sol beach; Oracle-L Subject: Re: Initialization parameter transactions_per_rollback_segment, can you set this? I would guess (the testing is up to you :) that the time saved during a restart of a job by committing each record does not come close to the time ( and frustration ) you would save by committing the whole batch once. -- Jared Still Certifiable Oracle DBA and Part Time Perl Evangelist On 8/16/05, Arnold, Sandra <ArnoldS@xxxxxxxx> wrote: It needs to commit after each record so that it will be easy to recover at the next record to be loaded if something happens during the load. For instance, if you are loading 100,000 records you do not need to load all 1000,000 again if a problem occurs during the load. You can pick back up at the next record to be loaded. The process logs each unique identifier that it loads so that you know where it was in the process when it failed.