Comparative Performance Measurement of the Pareto Optimal Combination and Multi-Objective Combination Models for Controller Placement in Software-Defined Networks

作     者：Mission Franklin Constance Izuchukwu Amannah Mission Franklin;Constance Izuchukwu Amannah

作者机构：Department of Computer Science Ignatius Ajuru University of Education Port Harcourt Nigeria 

出 版 物：《Journal of Computer and Communications》 (电脑和通信（英文）)

年 卷 期：2024年第12卷第3期

页      面：84-100页

学科分类：1304[艺术学-美术学] 13[艺术学] 08[工学] 0804[工学-仪器科学与技术] 0802[工学-机械工程] 

主　　题：Latency Measurement Metrics Performance POCO MOKO Architecture Provision 

摘      要：The evolution of the current network has challenges of programmability, maintainability and manageability, due to network ossification. This challenge led to the concept of software-defined networking (SDN), to decouple the control system from the infrastructure plane caused by ossification. The innovation created a problem with controller placement. That is how to effectively place controllers within a network topology to manage the network of data plane devices from the control plane. The study was designed to empirically evaluate and compare the functionalities of two controller placement algorithms: the POCO and MOCO. The methodology adopted in the study is the explorative and comparative investigation techniques. The study evaluated the performances of the Pareto optimal combination (POCO) and multi-objective combination (MOCO) algorithms in relation to calibrated positions of the controller within a software-defined network. The network environment and measurement metrics were held constant for both the POCO and MOCO models during the evaluation. The strengths and weaknesses of the POCO and MOCO models were justified. The results showed that the latencies of the two algorithms in relation to the GoodNet network are 3100 ms and 2500 ms for POCO and MOCO respectively. In Switch to Controller Average Case latency, the performance gives 2598 ms and 2769 ms for POCO and MOCO respectively. In Worst Case Switch to Controller latency, the performance shows 2776 ms and 2987 ms for POCO and MOCO respectively. The latencies of the two algorithms evaluated in relation to the Savvis network, compared as follows: 2912 ms and 2784 ms for POCO and MOCO respectively in Switch to Controller Average Case latency, 3129 ms and 3017 ms for POCO and MOCO respectively in Worst Case Switch to Controller latency, 2789 ms and 2693 ms for POCO and MOCO respectively in Average Case Controller to Controller latency, and 2873 ms and 2756 ms for POCO and MOCO in Worst Case Switch to Controll