Do We Need Resource Reservation?

Reservation reservation has been a very controversial issue among academic scholars and network operators. This paper examines the question of whether or not the Internet is better off with resource reservation from several aspects. An analytical model shows that reservation-capable networks provide better performance for real-time traffic than that in best-effort-only networks. A study on today’s Internet traffic volume, bandwidth pricing, and network topology indicates that network heterogeneity could become a significant problem that prevents ISP’s from delivering predictable services to end users. Finally, we look into some of the potential network applications, and conclude that supporting resource reservation is a key to the deployment of those applications. We don’t claim that we have given a definite proof on the necessity of resource reservation, and that the data we have collected represents today’s traffic condition in all ISP networks. However, we believe our results provide strong evidences that network over-provisioning has its limitation, and reservation protocols can be implemented and deployed with manageable impact on network’s scalability and on router’s overall performance. We hope that such evidence can foster more research on Internet reservation protocols.