Jointly Maximize Energy Saving and Congestion Avoidance in Hybrid SDN with Traffic Dynamics

Abstract

With the rapid growth of population and traffic in the Internet in recent years network devices join the Internet almost continuously in order to meet users' needs for network transmission As a result today’s networking hardware systems are required to run almost around the clock in order to keep the network running smoothly Consequently the global energy consumption of communication networks is extremely high The problem is exacerbated by the fact that the control plane and data forwarding plane are often embedded in the same device in current networks while network protocols are usually designed to operate in a distributed manner Thus with today's increasing focus on environmental issues devising the means to effectively reduce network energy consumption while simultaneously maintaining the performance of network communications is an extremely important but challenging problem Software Defined Networking is a newly-introduced networking paradigm in which the control plane is separated from the forwarding plane and moved to a globally-aware software controller By exploiting a centralized control mode the network traffic can be monitored in real time and rapidly rerouted as required to satisfy different objectives such as maximizing the network utilization achieving load balance and improving the energy efficiency However previous researches have focused only on the optimization of a single goal such as load balancing or energy saving Moreover they have generally ignored the dynamic characteristics of the traffic load in most networks Consequently a need still exists for effective strategies capable of simultaneously maximizing the network energy efficiency and preventing congestion avoidance under both light and heavy traffic load conditions Accordingly this thesis proposes a mechanism for reducing the energy consumption and avoiding network congestion in software-defined networks with a variable traffic load In addition a migration mechanism is proposed to minimize the energy consumption and congestion probability for the case where a traditional network is gradually upgraded to a software-defined network Through extensive simulations it is shown that the proposed mechanism has a good performance and reduces the energy consumption by around 40% on average under the premise of congestion avoidance Moreover the proposed migration mechanism also reduces the energy consumption by 40% on average while minimizing the probability of network congestion given around 60% of SDN nodes in the hybrid network

Date of Award	2017 Sept 1
Original language	English
Supervisor	Hui-Tang Lin (Supervisor)