Focusing on the issue of backbone bandwidth consumption, our simulation study demonstrated that connection forwarding results in noticeable bandwidth reduction. The reduction was 13% and 35% under the seven-day TTL assumption, depending on the policy used. The latency penalty was on the order of 10 milliseconds. Our experiments also showed that maintaining a dedicated output link at the content servers will greatly facilitate our mechanism, by improving the bandwidth savings from 13 to 35%, and at the same time make our approach more robust against different mechanisms for location information management.
We view our work presented in this paper a starting point for deploying forwarding requests mechanism within the context of cooperative proxies. We have already showed that this mechanism is useful in reducing web caching related backbone bandwidth consumption. In the future, we would like to investigate other potential benefits and to simulate the schemes on more extensive datasets. We also plan to refine our forwarding policy and implement a prototype system using Linux and Squid.