A Survey on Scheduling Techniques in the Edge Cloud: Issues, Challenges and Future Directions

TL;DR

This survey tackles the Edge Cloud scheduling problem, focusing on decisions about whether to execute tasks at the edge or the cloud under heterogeneous resources and mobility. It systematically classifies scheduling techniques into heuristic and meta-heuristic families, and further subcategorizes them by methods such as adaptive, QoS-based, ML-based, distributed, and prediction-driven strategies, including PSO, AFSA, and GA variants. The authors synthesize findings from 50 papers (2015–2021), compare QoS and cost metrics, and identify gaps—particularly in broad parameter coverage and machine learning approaches—toward guiding future work. The work aims to provide a rigorous, multi-faceted foundation for designing robust, scalable, and energy-efficient edge-cloud schedulers with practical relevance for real-world deployments.

Abstract

After the advent of the Internet of Things and 5G networks, edge computing became the center of attraction. The tasks demanding high computation are generally offloaded to the cloud since the edge is resource-limited. The Edge Cloud is a promising platform where the devices can offload delay-sensitive workloads. In this regard, scheduling holds great importance in offloading decisions in the Edge Cloud collaboration. The ultimate objectives of scheduling are the quality of experience, minimizing latency, and increasing performance. An abundance of efforts on scheduling has been done in the past. In this paper, we have surveyed proposed scheduling strategies in the context of edge cloud computing in various aspects such as advantages and demerits, QoS parameters, and fault tolerance. We have also surveyed such scheduling approaches to evaluate which one is feasible under what circumstances. We first classify all the algorithms into heuristic algorithms and meta-heuristics, and we subcategorize algorithms in each class further based on extracted attributes of algorithms. We hope that this survey will be very thoughtful in the development of new scheduling techniques. Issues, challenges, and future directions have also been examined.

Figures (6)

• Figure 1: The Edge Cloud computing scenario.
• Figure 2: Taxonomy of scheduling algorithms in the Edge Cloud.
• Figure 3: Machine Learning based techniques taxonomy.
• Figure 4: Supervised learning model process.
• Figure 5: Unsupervised learning model process.