Saving Energy in Partially Deployed Software Defined Networks
Huandong Wang,Yong Li,Depeng Jin,Pan Hui,Jie Wu
DOI: https://doi.org/10.1109/tc.2015.2451662
IF: 3.183
2015-01-01
IEEE Transactions on Computers
Abstract:As power consumption of the Internet has been growing quickly in recent years, saving energy has become an important problem of networking research, for which the most promising solution is to find the minimum-power network subsets and shut down other unnecessary network devices and links to satisfy changing traffic loads. However, in traditional networks, it is difficult to implement a coordinated strategy among the network devices due to their distributed network control. On the other hand, the new networking paradigm-software defined network (SDN) provides us an efficient way of having a centralized controller with a global network view to control the power states. As an emerging technology, SDNs usually coexist with traditional networks at present. Therefore, we need to investigate how to save energy in partially deployed SDNs. In this paper, we formulate the optimization problem of finding minimum-power network subsets in partially deployed SDNs. After proving the problem is NP-hard, we propose a heuristic solution to approach its exact solution. Through extensive simulations, we demonstrate that our heuristic algorithm has a good performance; that is, on average we can save about 50 percent of total power consumption in the full SDN, having a distance less than 5 percent of the exact solution's power consumption. Moreover, it also achieves good performance in the partially deployed SDN, on average saving about 40 percent of the total power consumption when there are about 60 percent SDN nodes in the network. Meanwhile, it runs significantly faster than a general linear solver of this problem, by reducing the computation time of the network containing hundreds of nodes by 100× at least.