A spatial HMM approach for green networking

Abstract

Green networking becomes more and more important, especially with the rapid development of data centers in recent years. In this paper, in order to minimize the energy consumption of a network, we present a novel energy saving approach, called Predictive Green Networking Approach (PGNA), based on a spatial Hidden Markov Model (sHMM). The sHMM is proposed to describe both the topology and the traffic distribution of the network. Loads of links in the network can be predicted based on the sHMM, and the links that most likely become near idle can be put into sleep mode to save energy. A deep-sleep method is proposed to maximize the energy saving while the network satisfies the connectivity and the maximum utilization constraints. We test the performance of PGNA with two real ISP backbone topologies and real traffic demands. The results show that our approach is effective and works better than related approaches.