Real-time Leader Election Algorithm in Asynchronous Distributed Systems

In this paper, a failure detection algorithm for efficient detection of failure occurred in some parts of a system consisting a process group in non-synchronized distributed system is proposed, and with applying the algorithm, a RTLE algorithm to make election of a new leader in case failure of the process working as a leader is also proposed. It is designed to make lists of currently working processes which are used for making list of next leader candidates in real-time. Each process references the list of candidates and the process with the highest priority works as a next leader in case of current leader failure. The evaluation results showed that proposed failure detection algorithm is more stable compared to A. Mostefaoui's algorithm even in environment with network delays. Proposed RTLE algorithm is also much more stable compared to other algorithms even in environment with changes in number of processes, network delay and number of additionally crashed leader candidates.