This paper addresses the concurrency control and buffer invalifation problems of Shared Disk DBMS. We show how both problems are solved by two decentralized integrated approaches. One approach is based on locking while the other relies on optimistic concurrency control.