Sorry for my misunderstanding. Innitially I thought the redo increase for small LOBs is caused by bug 6168063 but according to 66431.1 it is a feature. As for a bug then it is valid for 10.2.0.3 while we are on 10.2.0.4. Hope 10204 is not affected by it. But again, 66431.1 explains quite a lot about possible causes of flashback waits. Specially keeping in mind insufficient write performance (including redo and flash recovery area) we have on our storage. Thanks everyone for answers, Laimis N -- //www.freelists.org/webpage/oracle-l