On the Effectiveness of Optimization Search Strategies for Parallel Execution Spaces
Rosana S. G. Lanzelotte, Patrick Valduriez, Mohamed Zaït · Very Large Data Bases · 1993
The co.91 of query optimization is ufiected by both the ~eurch apocc und the search atmlegy of the opti,aizer. I~I (1 pumllel ecectrlion environment, Ihe search ap~cc tends lo he much lurger than in the centmlized ca,sc-. This is due to the high number of ezecution okeruutivrs which implies a aignijiconl increase in the op~imisutiofc coal. 118 &a puper, we investigate the tmde-o# between optimiaution cost and pumllel ezecution coat using the DBS$ pumllC1 query optimizer. We describe its cost model which cuptrrrea ull esaentiol aapects of pumllel executions. We show how the coal melricu imply a aignijiconl increase in the search spuce and oplimirotion coat. Howeuer, inateud of restricting lhe aeurch spabe, which muy lead to loosing letter plans, we reduce the optimization cost Iv controlling the search slmtegy. We ezrend mndomiaed strulegiea to adopt well lo pamllel query optimization. In particular, we propose Toured Simulated Annealing which provides u letter tmdec$ letween optimisa#on cost and yuulity of Ura ~~rullel r+rculiorr plan.