Improved Scheduling in Object Bases using Statically

Traditional approaches to scheduling in both object and data bases arrive at scheduling decisions based on dynamic, read/write behaviour or transaction submission order alone. Such approaches are based on exact but extremely limited information. Using static analysis, it is possible to derive more extensive but possibly inexact information. By combining the statically and dynamically derived information at scheduling time it is possible to increase the concurrency attained in transaction execution while decreasing the negative side-eeects of conventional concurrency control techniques including unnecessary overhead and the possibility of deadlock. In this paper, we describe a model and framework in which our approach may be applied. We then present the basic static analyses required to derive useful information for our concurrency control protocol and develop the scheduling algorithm itself. We conclude by arguing the eeectiveness of our approach and then brieey discuss future work being undertaken.