A Preference-Based Approach to Leader Election

点击翻译按钮获取中文摘要

A preference-based approach to the problem of leader election in distributed computer systems is examined. This approach is based upon the observation that a leader in a distributed system should represent a "good" choice for the system from a performance standpoint. Individual nodes "vote" for different candidate nodes based upon locally available information. The votes are combined using an election scheme to determine the leader. To develop and test this approach, it is planned to investigate its application to the problem of leader election in two specific system models: the delay model and the reliability model. Both, experimental and theoretical, techniques to develop and study different leader election schemes are used. The preference-based approach is of considerable practical interest and importance. Voting represents an efficient way by which to use locally available information to make a global decision that reflects the preferences of indididual nodes. Thus, the preference- based approach is not merely a solution to the problem of leader election; it represents, a generalized framework in which solutions to problems of choice may be developed.