Redundancy-Aware Routing with Limited Resources

Network load is reduced upon elimination of redundant data transfer. Redundancy elimination (RE) techniques can be applied on a per-packet basis, and provide benefit regardless of application. While it is straightforward to apply RE on a perlink basis, network cost can be further reduced by applying RE network-wide: by routing potentially redundant packets (identified using a redundancy profile) onto common links. Constructing redundancy-aware routes is challenging: it might not be economically viable to deploy RE over every link. Also, to preserve end-to-end performance and control signaling cost, routes cannot be determined on a per packet basis. We propose a redundancy-aware routing algorithm. Our approach can cope with limited resources (in terms of number of routers that can support RE ) and is feasible (not requiring for per-packet routing decisions). We evaluate our algorithm using detailed simulations, based on both synthetic traffic and trace captured from large enterprise networks. Unlike previous studies, our studies consider data from multiple sources. Our results show that a small number of RE equipped routers, coupled with our routing algorithm, are sufficient to achieve reduction in network load close to the unreachable upper bound.