Accurate prediction of operator execution time is a prerequisite for database query optimization. Although extensively studied for conventional disk-based DBMSs, cost modeling in mainmemory DBMSs is still an open issue. Recent database research has demonstrated that memory access is more and more becoming a significant-if not the major--cost component of database operations. If used properly, fast but small cache memories--usually organized in cascading hierarchy between CPU and main memory--can help to reduce memory access costs. However, they make the cost estimation problem more complex. In this article, we propose a generic technique to create accurate cost functions for database operations. We identify a few basic memory access patterns and provide cost functions that estimate their access costs for each level of the memory hierarchy. The cost functions are parameterized to accommodate various hardware characteristics appropriately. Combining the basic patterns, we can describe t...
Stefan Manegold, Peter A. Boncz, Martin L. Kersten