Reliability of Data Allocation on a Centralized Service Configuration with Distributed Servers

This paper presents a stochastic model to evaluate the reliability and availability of a network service configuration consisting of a centralized database server and two distributed database server sets for a local/metropolitan area network. The centralized server monitors and controls the distributed database servers, which can be remotely separated from each other and from the center by almost any distance depending on the means of information transmission media between the center and the distributed servers. Using exponential time to failure and time to repair distributions for each of the database servers, exact numerical solutions are obtained for the steady state service reliability and availability measures. The results are then compared to approximate steady state numerical solutions for the general Weibull distribution case under different initial conditions. Finally, the model is extended to the case of three distributed database server sets. The reliability and availability study of such configurations has many applications in widely used internet services, mission-critical business fault tolerant systems, and automated manufacturing and service industries. A decision tool for the computer center manager to assess an upgrade to a three distributed database system from a two database system is also presented.