RESOURCE MANAGEMENT ALGORITHMS FOR SOFTWARE DEFINED NETWORKS-BASED EDGE-CLOUD COMPUTING

Abstract

The integration of Software-Defined Networking (SDN), Edge Computing, and Cloud Computing represents a transformative synergy in modern network and computing architectures. SDN enhances network flexibility by separating control and data planes, a concept that becomes particularly valuable when combined with edge computing, which places computational resources closer to data sources. Cloud computing complements these advantages by offering scalable and on-demand resources to a wide range of applications and workloads, and ensuring resource availability across the network. Recent advancements consider the adoption of SDN infrastructure to empower cloud and edge computing for dynamic controllability and manageability. However, the integration of SDN into cloud and edge poses key challenges, including suboptimal resource utilisation in heavily-loaded SDN-Cloud networks, which leads to network congestion, QoS violations, and increased power consumption. Additionally, controller congestion in SDN systems leads to delays, reduces scalability, and prevents the system to handle high traffic loads efficiently, posing a significant challenge for optimising network performance. While conflicts in prioritisation complicate the efficient allocation of resources, which can degrade QoS and network efficiency. To address these challenges, three algorithms are proposed for SDN-Cloud and SDN Edge-Cloud platforms. The Dual-Phase Virtual Machine (VM) allocation algorithm (D-Ph) optimises resources in SDN-Cloud networks, considering processing capacity and memory requirements, to enhance QoS and power efficiencies. The Queue Theory Model-based Adaptive Reinforcement Learning Algorithm (QTM-ARL) optimises load balancing in SDN Edge-Cloud platform while maintaining QoS constraints. Priority-Aware Scheduler (PASQ) based on QoS constraints and incorporated with rate limit mechanism, manages network traffic efficiently while prioritising VoIP traffic over video streaming to enhance network performance. The proposed algorithms are investigated for performance through eventdriven simulation (CloudSimSDN) and MATLAB, employing real workload datasets and delay-sensitive applications. Results demonstrate D-Ph's efficiency in balancing network performance and power consumption in heterogeneous heavily-load large-scale SDN-Cloud networks based on response time, network and CPU performance, QoS violation rate, and power consumption. Furthermore, QTM-ARL's effectiveness in maintaining QoS in hierarchical multi-controller system with fluctuating data flows, and PAS-Q's ability to prioritize low-latency VoIP traffic over video streaming while achieving the desired level of service quality for real-time communication applications and thus fair resource utilisation. Future research can explore advanced AI, emerging technologies, eco-friendly practices, and adaptive SDN architectures to enhance the efficiency, security, and sustainability of SDNbased Edge-Cloud systems.