QUALITY OF SERVICE OPTIMIZATION IN HYBRID SOFTWARE- DEFINED NETWORK

Abstract

Traditional network infrastructures offer numerous features but exhibit limitations, particularly in providing advanced network control for implementing novel concepts. For future network design, different architectures worldwide have been proposed; among them, one of the finest network designs is Software-Defined Networking (SDN), which separates the network control and data layers, offering enhanced agility, programmability, flexibility, and advanced traffic engineering capabilities. However, because of economic and technical challenges, directly upgrading the entire conventional network to fully SDN is challenging for many organizations. The optimal solution to get the advantages of both networks is to incrementally upgrade conventional network devices to SDN nodes, called the Hybrid SDN network. The amalgamation of SDN principles with conventional networking techniques has led to the emergence of hybrid SDN; as the current network infrastructure evolves, it combines the programmability of SDN with the traditional protocols of conventional networking systems. A practical approach involves adopting a hybrid SDN model, wherein conventional and SDN components are integrated seamlessly. This paradigm change presents new challenges and opportunities in Traffic Engineering (TE), demanding novel approaches to enhance network performance, optimize resource usage, and increase overall efficiency. This thesis aimed to investigate the existing literature on the migration sequence to a hybrid SDN and traffic engineering optimization in the hybrid SDN environment. Specifically, we examined the focus of most authors, which primarily revolved around minimizing Maximum Link Utilization (MLU) in the hybrid SDN environment. Therefore, we introduced more precise methods to determine the optimal migration sequence toward a hybrid SDN architecture, and further refined both the OSPF weight configuration and the SDN nodes' traffic splitting ratios by incorporating a wider range of parameters in the prediction process. To address the research gaps, we defined four primary objectives. The first objective was to review and compare the existing methods implemented for the migration from a conventional network to a hybrid SDN network, and then traffic engineering optimization in the hybrid SDN environment. The second objective was to propose a model for the migration of the conventional legacy network to the hybrid SDN network. We introduced the FCM (Flow Control Method) approach, which investigates the gradual deployment of SDN nodes in conventional networks using a simple greedy algorithm, presenting a potential solution for the evolution of network architecture. We investigated the optimal migration sequence from a conventional network to a hybrid SDN to address the optimization impact on controllable traffic. The proposed technique efficiently integrates SDN features while minimizing network disruption and optimizing link utilization. We evaluate the optimal deployment of SDN nodes v through simulations, considering network topologies, available resources, and traffic patterns. Simulation experiments conducted on real network topologies demonstrate that by upgrading 17% of the nodes to SDN, near-optimal performance can be achieved while requiring substantially less investment in resources and effort for network upgrades. This result suggests that the greedy method is cost-effective and highly effective in practical scenarios. Achieving near-optimal load balancing with minimal upgrades could offer significant advantages for network operators, making it a valuable strategy for managing network performance in real-world applications. The third objective entailed a mechanism for routing optimization in the migrated hybrid SDN network. In a hybrid SDN network, where traditional and SDN devices operate simultaneously, the transition introduces new challenges and opportunities for traffic engineering. This demands innovative approaches to enhance network performance, resource utilization, and overall efficiency. We explored routing optimization for traffic engineering within this evolving hybrid environment and addressed the problem as a mixed-integer nonlinear programming (MINLP) framework. We introduced a heuristic technique, H-STE, to improve traffic engineering in the hybrid SDN; we concentrated on minimizing the MLU by optimizing two critical aspects: optimizing the OSPF weight settings across the whole network to balance the flows originating from conventional devices and optimizing the traffic splitting ratio of SDN nodes. The H-STE method helps bridge the gap between the two by optimizing the whole network performance and adhering to the limitations imposed by legacy routing infrastructure. We have conducted several experiments on real network datasets; the finding shows that with a 30% deployment ratio of SDN nodes, we can reduce the MLU and get close to the optimal result to get maximum benefits from the concept of the hybrid SDN. This research contributes to the evolution of communication infrastructure by promoting efficiency, flexibility, and robustness in the pursuit of better network management and design. The fourth objective was to develop a framework for the interconnection of SDN nodes alongside conventional network devices. We investigated the interconnection of SDN nodes with conventional legacy devices in a hybrid SDN network. We studied the integration of SDN nodes alongside conventional legacy networks and explored how to peer and integrate SDN nodes with conventional legacy networks using the SDN-IP application located on the application layer of the ONOS controller. As multiple SDN controllers exist, selecting the appropriate controller based on the network topology is the key to achieving optimal network performance. We have studied and examined different controllers for the hybrid SDN environment, as multiple controllers exist. Notable controllers are NOX, Floodlight, POX, ODL, Ryu, and ONOS. Finally, the ONOS (Open Network Operating System) controller was chosen as the SDN controller for our hybrid SDN setup due to its compatibility with hybrid SDN networks and its support for SDN-IP applications. Nevertheless, none of the earlier studies interconnect SDN nodes with conventional legacy networks using ONOS controllers in a real network scenario. We vi have considered this issue as the interconnection of SDN nodes into conventional legacy networks with ONOS controllers using Mininet tools. Through extensive simulation, we found that hybrid SDN is the optimal solution for organizations seeking a modern network infrastructure without network disruption and service downtime, a hybrid SDN can provide a network environment that is more scalable, adaptable, and programmable. By addressing these objectives, our research offers valuable contributions to network performance enhancements in the hybrid SDN environment. The scalability of our approach ensures that the method maintains its efficiency as the topology expands, demonstrating its resilience in controlling diverse network topologies. This is a significant improvement as it reduces resource utilization while enhancing network efficiency.