Skip to Main Content
The Valiant Load-Balancing (VLB) design has been proposed for a backbone network architecture that can efficiently provide predictable performance under changing traffic matrices . In this paper we show that the VLB network has optimal performance when nodes can fail, in the sense that it can support the maximal homogeneous flow for any number of node failures. We generalize the VLB design to enable interconnection of multiple VLB networks, and study interconnection via bilateral peering agreements as well as transit agreements. We show that using VLB as a transit scheme yields the lowest possible network and interconnection capacities, while VLB peering can also achieve near-optimal use of capacity.