Broker based Framework for Service Orchestration in Cloud and Fog Computing Environments

Abstract

Distributed Systems have significantly contributed to the evolution of the field of computing by enabling job distribution and resource sharing. Distributed computing paradigms such as Cloud and Fog have eased users by making high-end resources available to the user without even possessing them physically. Users of these computing paradigms can use resources by paying nominal costs per their usage, scale their subscription, or shift from one set of resources to another as per their changing computational requirements. In this scenario, efficiently handling the requests and resources becomes crucial so that the available infrastructure can be utilized optimally. Job scheduling and load balancing are the formal ways to map the job requests on appropriate resources in DS appropriately. Both job scheduling and load balancing are prevalent research topics among researchers, and a considerable volume of literature is available to introduce tools and techniques for it. These techniques are expected to resolve the challenges faced by distributed computing, such as imbalanced load, the geographical coverage of resources, over/under utilization of resources, meeting Service Level Agreement, maintaining Quality of Service, and curbing the number of denied jobs. This thesis presents three frameworks: the first is a software agent-based broker framework for load balancing in a Partitioned Public Cloud, an ML classification-based dual broker framework that integrates QoS parameters for scheduling applications’ jobs in FEs. This framework categorizes the resources available at the nodes as compute-intensive, memory-intensive, and GPU-intensive.