PLASTIC [1] is a recently-proposed tool to help query optimizers significantly amortize optimization overheads through a technique of plan recycling. The tool groups similar queries into clusters and uses the optimizer-generated plan for the cluster representative to execute all future queries assigned to the cluster. An earlier demo [2] had presented a basic prototype implementation of PLASTIC. We have now significantly extended the scope, useability, and efficiency of PLASTIC, by incorporating a variety of new features, including an enhanced query feature vector, variablesized clustering and a decision-tree-based query classifier. The demo of the upgraded PLASTIC tool is shown on commercial database platforms (IBM DB2 and Oracle 9i).
Parag Sarda, Jayant R. Haritsa