A distributed system is a collection of computers that are connected via a communication network. Distributed systems have become commonplace due to the wide availability of low-cost, high performance computers and network devices. However, the management infrastructure often does not scale well when distributed systems get very large. Some of the considerations in building a distributed system are the choice of the network topology and the method used to construct the distributed system so as to optimize the scalability and reliability of the system, lower the cost of linking nodes together and minimize the message delay in transmission, and simplify system resource management. We have developed a new distributed management system that is able to handle the dynamic increase of system size, detect and recover the unexpected failure of system services, and manage system resources. The topologies used in the system are the tree-structured network and the ring-structured network. This thesis presents the research background, system components, design, implementation, experiment results and the conclusions of our work. The thesis is organized as follows: the research background is presented in chapter 1. Chapter 2 describes the system components, including the different node types and different connection types used in the system. In chapter 3, we describe the message types and message formats in the system. We discuss the system design and implementation in chapter 4. In chapter 5, we present the test environment and results. Finally, we conclude with a summary and describe our future work in chapter 6.