Energy Aware Traffic Engineering with Reliability Constraints

Abstract

Current network infrastructures are over-provisioned to increase their resilience against resource failures. Such strategies exhibit poor energy efficiency during off-peak periods. In this respect, energy aware Traffic Engineering (TE) solutions aim to maximally switch off redundant network resources when traffic load is low. However, these green TE solutions do not consider their effects on network fault-tolerance. In this paper, we first aim to quantify the effects of five recently proposed green routing approaches, namely FGH, GreenTE, MSPF, SSPF, and TLDP, on the following two reliability measures: (i) terminal reliability (TR), and (ii) route reliability (RR). Experiments using three topologies with real and synthetic traffic demands show that green approaches that switch off redundant links affect TR and RR significantly. Specifically, routing traffic through multiple paths impacts reliability less while reducing energy, especially when the paths are link disjoint. Interestingly, TLDP and MSPF have better route reliability than using shortest path (SP) routing. We then formulate a problem, called reliable energy-aware-routing (R-EAR), which aims to maximally switch-off network cables subject to link utilization as well as TR/RR requirement. We also propose an effective algorithm, called reliable Green Routing (R-GR), to solve R-EAR. Evaluation on three real topologies shows that R-GR can save energy while satisfying both reliability and link utilization requirements. Specifically, for the GEANT network, R-GR saves up to 25.65% in energy usage without reducing RR, and saves up to 48.65% with an average RR reduction of 9.1%. For the same network, our solution achieved an energy saving of 28.38% while reducing the average TR by only 8.5%.