Today I was on The Heap and was looking at a query plan for something I thought could be improved. However, it created something that shook my belief in the SQL Server query optimizer. Can I still trust sql-server if it cannot even count to 100%?
Characteristics of the table:
- clustered on a non-identity column
- 12 indexes, one of them being the
date_enteredcolumn in question
- 60,000 records
- 26 columns of various types and length
- PAGE compressed table
Has anyone seen this before and what causes the plan to look so skewed?
And below from SQL Sentry Plan Explorer