Parallel queries incur more overhead costs to perform such internal tasks as:
Allocating and initializing worker processes
Coordinating worker processes as they execute a query plan
Deallocating worker processes after the query is completed
To avoid applying these overhead costs to OLTP-based queries, the optimizer “disqualifies” tables from using parallel access methods when a scan would access fewer than 20 data pages in a table. This restriction applies whether or not an index is used to access a table’s data. When Adaptive Server must scan fewer than 20 data pages, the optimizer considers only serial table and index scans and does not consider parallel optimization.
