A Pre-committed Concurrency Control Scheme In Mobile Database System

In last ten years, the improvements and integration of computing technology and wireless communication made a new computing mode, mobile computing, become reality. The research on database management in mobile environment, which is known as mobile computing, has become a new hot topic in database reseach society. Research mainly focuses on replication and caching, location management, query optimizing, data broadcast, security, mobile transaction processing and human-machine interface technologies. Among these topics, research of mobile transaction processing is one of the key technologies, while the study of concurrency control is the most important part of mobile transaction processing.Among the traditional schemes of concurrency control, for instance 2PL, the optimistic concurrency control scheme, the time-compositor concurrency control scheme, are all cost much in the way of setting lock and examining the conflicts. So these methods don't adapt to the mobile environment. In this paper, we study a new model of mobile database which is based of the two-tier replication method and the classical model of the mobile database. This new model includes two parts: the fixed network and the mobile network. The servers of the mobile database system compose the fixed network. This corresponds the first replication of the two-tier replication, the replication of the servers. To improve the efficiency of the first replication, we use the technology of P2P and the DRG Also, the mobile network is consitituded of the mobile clients. The mobile clients communicate with the nearest server through the wireless network. This is the second replication of the two-tier replication method.The second replication uses the pre-committed concurrency control scheme. This scheme allows some of the mobile clients to pre-commit at the local place when the link is disconnected, but the other of MCs must wait until the link is re-connected. To confirm which MCs can pre-commit, we should choose the value of the amount of change that can occur on the replica at each MC (we keep it asλ). If the transactions change the data value by at mostλin a MC, they are called pre-committed transactions and can be free to commit. Otherwise, if the transactions change the data value more thanλ, they are called request transactions and must commit at the server.Experiment results of the DRG and the pre-committed scheme indicate that the methods can reduces transactions' commit time and also help to reduce the number of rejections, which could arise due to the BT not being able to commit.