Yuki Kusumi

File access level optimization using page access graph on recursive query evaluation

As the performance measure for recursive query evaluation algorithms in the deductive database system, the largest amount of necessary storage space for intermediate results measured by the number of tuples has been usually employed. In this paper, using a practical performance evaluation measure based on the physical file access; namely, the number of pages accessed, we will investigate a general framework of efficient file access strategies for the transitive closure computation, which is one of performance bottlenecks in the recursive query evaluation. We introduce the page access graph for representing the page structure of a given file, and this graph provides the basis for considering the optimization of page access scheduling. Under the assumption that whole or a part of this graph is processed in the main memory, if the graph is acyclic, our algorithm provides the optimal page access scheduling, i.e., each page has to be fetched at most once.