What causes the sort order of parallel query results to differ from time to time?
A query that neither includes aggregates nor requires a final sorting step returns the same set of results each time; but the sort order of the results may vary as follows: • A parallel access method may return results in a different order than a serial access method does. • The query, run in parallel at different times, may return results in a different sort order each time. For a parallel query, each worker process sends results to the network buffer, and then the coordinating worker process (request originator) merges the results of all the worker processes. The worker process speeds differ, depending on such factors as already-cached pages and lock contention; and the order in which each worker process returns a result set impacts both the merge and the order of the final result set. Types of queries that may return results differently when there is no aggregate or final sort include: • Queries that select all or a large number of rows. The result set is the same but the order is d
