Current database systems were designed assuming that data resides on disk. This assumption is no longer valid; main memories have become sufficiently large that most OLTP databases can reside entirely in memory. In this project we investigate what architectural changes and new techniques are required to realize the potential for great performance improvements offered by storing data in main memory.