Abstract: The power consumption of Internet is increasing year by year. How to decrease the consumption is of interest for both the research community and the public. Current network has poor efficiency because of the low link utilization and high path redundancy. The low link utilization and high path redundancy lead to high energy consumption. In order to decrease the energy consumption and improve the efficiency, this paper proposes an EAR algorithm called TM-EAR which is combined with the traffic matrix to improve the link utilization and reduce the number of path redundancy. TM-EAR can find a spanning tree which all traffic data are routed on. And we can switch off the links that are not contained in the spanning tree. The experiments show that it is possible to switch off about 50-20% of links. The TM-EAR has a good performance.