Enhanced Particle Swarm Optimisation and Multi Objective Optimization for the Orchestration of Edge Cloud Clusters

Hafiz Shahid, Claus Pahl


Load balancing and workload distribution cause challenges for the management of IoT and distributed systems in the edge computing environment. Swarm intelligence is a technology suitable for the management of distributed systems, networks, communication and routing protocols. Swarm intelligence-based PSO algorithms (particle swarm optimization) can be applied for load balancing and task scheduling in cloud computing environments operating through a broker agent. In distributed cloud environments, data is collected and then processed at the center of the cloud, rather than making decision at edge nodes closer to IoT infrastructures. Here, we develop an automated orchestration technique for clustered cloud architectures. An Autonomous Particle Swarm Optimization, called the A-PSO algorithm, is implemented that enables an edge node, such as a remote storage, to work as part of a decentralized, self-adaptive intelligent task scheduling and load balancing agant between resources in distributed systems. Using Multi Objective Optimization (MOO), complementing the A-PSO algorithm, we also include metrics such as Actual Round-Trip Time (ARTT) of tasks assignments to the remote storage to reduce the execution cost. Our A-PSO algorithm can orchestrate the distribution of large volumes of data to remote storage and back in cluster, i.e., coordinated distributed cloud environments.


Paper Citation