A fast commit protocol for distributed main memory database systems

Abstract

Although the distributed database systems has been studyed for a long timeÿ there has been only few commercial systems available. The main reason for this is that the distributed commit processing costs too much which results in little or no performance gain compared with single node database system. In this paperÿ we note the di÷erence in the update and logging policy between disk based database and main memory database in the distributed environmentÿ and presents a fast distributed commit protocol for the main memory database. In the proü posed protocolÿ instead of sending and receiving two sets of messages one after the other as in two phase commitÿ only one set of messages are sent after the coordinator completes committing a distributed transaction. The main idea of this fast commit processing is to send all the redoülogs to the coordinator so that the coordinator alone can make the decision to commit or abort when the time comes. As a resultÿ the frequency of the communication and the disk access related to the commit processing can be signiöcantly reduced. Our simulation study shows that the proposed commit protocol achieves the high performance as we expect.