Parallel join processing using nonclustered indexes for a shared memory multiprocessor

Examines the problem of efficient join processing on a shared memory multiprocessor, when the available access paths that are chosen (by the query optimizer) are nonclustered indexes on the joining attributes. Two algorithms based on simple heuristics and a third algorithm based on the Monte Carlo optimization technique called simulated annealing are given. Experiments are performed to compare the various algorithms with an algorithm that recently appeared in the literature