Database applications are rapidly emerging as important candidates for parallel solutions, especially in distributed environments. We present a strategy for efficient transaction management in an environment of replicated autonomous databases. Two such protocols are developed, one for a LAN environment and one for remote replcation involving lower communication overhead but assuming a loose logical partition of the database schema. Our protocol improves response time and throughtput by exploiting parallelism at the expense of reducing the level of transaction isolation. We are able to prove that users of the parallel replicated system can view it as equivalent to that of a single database providing `degree 2' transaction isolation, i.e. the replication and parallelism is transparent to the application programmer. Experimental results using a prototype demonstrating potential performance improvements are also presented.

Back to Workshop Schedule

1996-1997 Mathematics in High Performance Computing