Design of multiple reverse spanning trees in next generation of ethernet-VPNs

Abstract

Recently, a new global open Ethernet (GOE) architecture was proposed as a cost-effective scalable solution for the next generation of VPNs over an optical network infrastructure. In this paper, we present a traffic engineering design for networks that support the GOE architecture. The current extensions of the spanning tree protocol (STP), such as rapid spanning tree (RST) and multiple spanning tree (MST), use a link cost metric which is inversely proportional to the link capacity. In order to achieve a better load balancing in the network, we propose to use the link cost metric which is a function of the current link load. The traffic engineering algorithms presented in this paper use two different link cost functions. We comment on the performance of those heuristic algorithms for various network topologies and under several traffic scenarios. Our results demonstrate a significant improvement in the network utilization when the link cost metric is a function of the link load compared to the case when it is inversely proportional to the link capacity.