Dynamic Data Replication in Distributed Systems

Yixiu Huang

1995-01-01

Abstract:We are presently moving towards a distributed, wholly interconnected information environment, in which data items will be accessed from many locations that may be geographically distributed world-wide. In such an environment, data items are usually replicated, i.e., stored in local databases of multiple processors, for data access availability and its performance. Improper management of the replicated data may result in heavy network traffic, and lead to severe performance problems. Such a problem is especially significant in the mobile computing environments, in which the aerial bandwidth is limited, and wireless communication is much more costly than the cable wired network communication. In this thesis, several mathematical models for evaluating the network traffic involved with the replicated objects are presented. Convergence is introduced for evaluating the replicated data management algorithms of regular network traffic. Competitiveness is introduced for the worst case analysis. A stochastic method is introduced for the expected-case study. Meanwhile, I will present a couple of dynamic data replication algorithms, each of which is for different network architecture. I will compare the proposed algorithms with the traditional static management algorithm for the replicated data, analytically and experimentally. The analytical comparison is based on the above proposed measurements, and the experimental comparison is based on our simulation results.