Greening data center networks with throughput-guaranteed power-aware routing

  • Authors:

  • Mingwei Xu;Yunfei Shang;Dan Li;Xin Wang

  • Affiliations:

  • Department of Computer Science and Technology, Tsinghua University, Beijing, PR China and State key Lab. of Networking and Switching Technology Beijing University of Posts and Telecommunications, ...;Department of Computer Science and Technology, Tsinghua University, Beijing, PR China;Department of Computer Science and Technology, Tsinghua University, Beijing, PR China;Department of Electrical and Computer Engineering, State Univ. of New York at Stony Brook, NY, United States

  • Venue:
  • Computer Networks: The International Journal of Computer and Telecommunications Networking
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

Cloud based applications and services require high performance and strong reliability provided by data center networks. To overcome the problem of traditional tree based data center network, recently many new network architectures are proposed, such as Fat-Tree and BCube. They use aggressively over-provisioned network devices and links to achieve 1:1 oversubscription ratio. However, most of the time data center traffic is far below the peak value and a large number of idle network devices and links in data centers consume a significant amount of power, which is now becoming a big problem for many cloud providers. In this paper, we aim to reduce the power consumption of high-density data center networks from the routing perspective while meeting the network performance requirement. We call this kind of routing throughput-guaranteed power-aware routing. The essence of our idea is to use as little network power as possible to provide the routing service, without significantly compromise on the network performance. The idle network devices and links can be shut down or put into the sleep mode for power saving. Extensive simulations conducted in typical data center networks show that our power-aware routing can effectively reduce the power consumption of network devices, especially under low network loads.