I maintain a database serving as a backend for a vendor supplied application purchased by my department. The majority of activity (90%) consists of update/insert statements against a single table that contains 14 million rows (pointers to images on an optical jukebox). The table has 4 indexes defined, but two of them(non-clustered) have only two distinct values one of which is NULL. With such low cardinality, I doubt the optimizer will ever make use of them and considering the type of activity (80% update/insert) that's a chunk of overhead for something that won't be used. The users are complaining about system performance. I captured the update and occasional select statements being issued and their execution plan indicates that they are accessing the table using an index seek on the clustered index defined on the table in question. Can anyone think of a reason to keep the other two indexes around?