Affiliations: School of Electronic and Electrical Engineering, University of Leeds, Leeds, UK | Department of Electrical and Computer Engineering, King Abdulaziz University, Jeddah, Kingdom of Saudi Arabia
Note: [] Corresponding author: Niemah I. Osman, School of Electronic and Electrical Engineering, University of Leeds, Leeds, LS2 9JT, UK. E-mail: elnio@leeds.ac.uk
Abstract: Increasing popularity of media-rich Internet content and the associated growth in power consumption have highlighted the need for energy efficient content distribution schemes. This paper investigates the power savings introduced by caching content in IP over WDM networks. To carry out the study, we develop a MILP model to minimize the power consumption of cache-based services over an IP over WDM network by optimizing the cache sizes of the network nodes at different times of the day. A Constraint-Based Genetic Algorithm (CBGA) is developed to validate the optimum cache sizes obtained from the MILP model and a simulation based on lightpath bypass validates the power consumption of routing traffic demands. We consider different popularity distributions including the Zipf, Pareto and Bimodal content popularity distributions to exemplify different types of IPTV services. Our results reveal that deploying the optimum variable cache sizes at the nodes reduces the overall network power consumption by up to 42%, 68% and 72% using Zipf, Pareto and Bimodal distributions for content popularity, respectively. We further study the impact of regular cache updates on power efficiency. The results show that removing the 10 most popular objects from caches increases the network power consumption by up to 20%.
Keywords: Cache hit ratio, content popularity distribution, IP over WDM
