Figure 24-2 shows how parallel query processing reduces response time over the same query running in serial. In parallel execution, three worker processes scan the data pages. The times required by each worker process may vary, depending on the amount of data that each process needs to access. Also, a scan can be temporarily blocked due to locks on data pages held by other users. When all of the data has been read, the results from each worker process are merged into a single result set by the coordinating process and returned to the client.
Figure 24-2: Relative execution times for serial and parallel query execution
The total amount of work performed by the query running in parallel is greater than the amount of work performed by the query running in serial, but the response time is shorter.