The main responsibilities of the relational engine are:
The parser scans an SQL statement and breaks it down into the logical units, such as keywords, parameters, operators, and identifiers. The parser also breaks down the overall SQL statement into a series of smaller logical operations.
Typically, there are many ways that the server could use data from the source tables to build the result set. The query optimizer determines what these various series of steps are, estimates the cost of each series (primarily in terms of file I/O), and chooses the series of steps that has the lowest cost. It then combines the specific steps with the query tree to produce an optimized execution plan.
After the query optimizer has defined the logical operations required to complete a statement, the relational engine steps through these operations in the sequence specified in the optimized execution plan.
These statements are not the typical SELECT, INSERT, UPDATE, or DELETE statements; these statements have special processing needs. Examples are the SET statements to set connection options, and the CREATE statements to create objects in a database.
The relational engine formats the results returned to the client. The results are formatted as either a traditional, tabular result set or as an XML document. The results are then encapsulated in one or more TDS packets and returned to the application.