Dynamic offloading in a multi-provider environment: A behavioral framework for use in influencing peering

Abstract

We pose the question of how to encourage the resource sharing in a distributed, multi-provider environment, where each node, or provider, has local work but is able to accept additional work from other nodes/providers if there is available capacity. An instance of such an environment is found in content delivery, where numerous, competing providers can work together if enough benefit is to be gained from doing so. We model individual provider behavior as essentially selfish, and then propose pricing schemes to exploit the selfishness to achieve system wide performance gains. We employ a game theoretic framework to analyze the problem, and come up with a time-dependent, non-cooperative network equilibrium model. To influence the system towards the positive end of resource sharing, we suggest the creation of a monetary unit, tokens, whose exchange encourages a more efficient use of system-wide capacity, and whose effect is regulated by the pricing scheme in place. The impact of the different node behavior, model parameters, and pricing schemes in influencing the system performance is investigated through simulation. This framework can be combined with distance and round trip time to calibrate redirection behavior of distributed server environments.