OpERA: A Multi-Objective Optimization Approach for Edge Based Resource Allocation

Abstract

In the Internet of Things (IoT) era, as system complexity increases and IoT systems require greater complexity, adaptive optimization strategies are required to improve computation offloading across computing infrastructures. For most offloadable tasks, offloading is targeted to the cloud, in recent years efforts have been made to migrate away from the reliance on the cloud because of the pitfalls it presents for latency-sensitive applications. Solutions proposed include task preprocessing or computation offloading to multi-access edge computing nodes. However, those solutions come with challenges related to complex hardware and software requirements, expansive implementation and complex architecture. As part of this thesis, we propose OpERA, an edge-based resource allocation optimization framework for aiding in seamlessly offloading tasks across edge, fog, and cloud computing layers and architectures. By capturing the task attributes and requirements from the end user, OpERA is able to identify suitable resources to execution the task and then optimizes the process by singling out the computational resource with the best residual gains (i.e., cost reduction, minimal energy consumption, high processing capability, etc.). By optimizing resource allocation in computation offloading in the status quo, we are able to increase the likelihood of successful task offloading for computationally intensive tasks such robotic surgery, autonomous driving, smart city monitoring device grids, and deep learning tasks.