Study On The Management Method Of Distributed Transaction Based On Redundancy And Cooperation

With the development of computer technology and the popularity of internet network, distributed transaction management becomes more and more important. Its primary target is to maintain the properties of the distributed transaction, such as Atomicity, Consistency, Isolation and Durability. To achieve this target, a commit protocol is adopted to manage transaction. It is desired that the commit protocol has the characteristic of higher reliability and lower overhead.Firstly, this paper analyzes the prevalent distributed transaction commit protocols in detail, including Two-phase commit(2PC) and Three-phase commit(3PC). 2PC protocols can achieve lower overhead, but they are blocking. In contrast, 3PC protocols can achieve nonblocking, but cause more overhead. 3PC protocols permit each site to proceed in its operation, without waiting for recovery of the failed site. However, participants need to concern the issue of voting a new coordinator in a 3PC protocol, thus bringing extra overhead.Secondly, this paper proposes a new revision to 3PC protocol based on coordinator redundancy, in which the participants believe that the coordinator is fault-free by constructing redundancy of coordinator . In the revised protocol, the participants need not consider to vote a new coordinator, lowering the overhead of the original 3PC. Because of existing a backup of coordinator, the coordinator's failure can be recovered quickly.In the system, rollback recovery technique using checkpoint is adopted for fault tolerance, bringing some extra overhead. In the rollback recovery using checkpoint, proper determination of checkpoint interval can make system performance optimized. This paper derives an equation about the best checkpoint interval from Markov chain in the duplicated fault tolerance system, and testifies its correctness through the experiment.In the end, the performance of system is evaluated through the experiment involving nanotube molecular devices. The evaluation shows that the revised system can achieve higher reliability and fewer overhead , and can make fault recover quickly.