Dynamic resource management algorithms for edge computing using hardware accelerators

Abstract

Many Internet of Things (IoT) applications must perform their Big Data analytics tasks at the edge to meet their real-time needs and overcome the constraints on and reliability of network resources. Since traditional CPUs cannot meet these demands, solutions are sought by using accelerator hardware such as FPGAs, GPUs and TPUs to address these challenges. My doctoral research is focusing on ascertaining the feasibility of utilizing these accelerators for real-time IoT Big Data analytics, and in turn investigating dynamic resource management algorithms to schedule edge-based accelerator resources in the presence of highly dynamic IoT workloads.