Design and Implementation of a Fault Tolerant Multiple Master Cloud Computing System

Today the applications of computing machines have outreached domains such as quantum physics, weather forecasting, or molecular biology, to name a few. Distributed computing systems provide a solution to meet the huge computational requirements by combining the computational capacity of multiple computing machines. However, most of the current systems are based on Single Master Multiple Slave (SMMS) architecture, which has limitations in terms of scalability, and fault management because of its localized administration and physical resource constraints. This paper introduces a novel general-purpose and fault-tolerant distributed computing architecture (so-called Multiple Master Multiple Slaves - MMMS) that can be highly scalable, and efficient on a non-memory shared model. In the paper, a dynamic algorithm for creation and management of MMMS architecture is presented which enhances the fault tolerance capabilities of the system.