I executed an update on a single column in a table with 225 million records via a scheduled job. After 3 hours, I decided to kill the session and modify the query to try and get it to run faster. The subsequent rollback is estimated to take 35 hours (select used_urec from v$transaction;). It doesn't seem like a query should take 10 times longer to rollback. Is it possible that one of the database parameters is misconfigured? The SGA is approximately 20 GB.