Scalable Leader Election Considering Load Balancing

Abstract

Distributed computing consists of a model where multiple parts of a system are accessed by different computing machines for the betterment of work efficiency. To get better results, electing a leader is one of the most critical tasks. Leader election is the process of assigning one of the processes as the leader whose work is to organize all the provided jobs that are distributed between the different computer nodes and to provide them with the required resources. The obstacle that is dealt with here is to elect a processor that can act like a leader from among the set of multiple processors using distributed protocols. To explain this in further detail, let us consider that we have n number of processors, and among these n processors, we have n number of processors, and among these n processors we have cn number of processors which can be considered as bad or corrupt and (1-c)n number of processors among them can be considered as good or not corrupt. Here the value of c is a fraction value and is fixed. The problem that is to be dealt with is to select a processor with a probability that has to be constant, a single processor from the given n number of processors which can act as their leader, no matter which set of the given cn processors are bad here. The scalability that is mentioned hereof leader election being scalable refers to the fact that every good or non-corrupt processor that is available amongst the total n number of processors sends and also processes several bits. And these number of bits that are being sent and processed by the various processors are polylogarithmic in n. Here we can say that the number of bits that are sent over or processed by a node belongs to a function that is polynomial in the logarithm of n.