|
Even though the query optimizer advises creating an index, it doesn't always choose to use it when creating its access plan. One reason for this is that the query optimizer does not always advise the perfect index. This is because currently, only local selection is used when determining what index to advise - it does not yet consider join, order by, or group by columns. As a result, after you create that advised index, it may not choose to use it, as it is still not the perfect index. You may need to manually intervene - examine the query and consider the join, group by, and order by columns to determine what the perfect index is and create that index.
Alternatively, it may be recommending that index so that it can extract stats from the index to further evaluate the optimal access method. For example, once the index has been created, it may be using statistics from the index to determine that that query will return many rows, and thus a table scan is the least-expensive access method.
To better understand the concept of a perfect index and how indexes help the query optimizer with statistics, you can read the Indexing Strategy white paper here.
|
