From multi-description/multi-path routing for multi-media applications to content distribution in P2P networks to community networking, many forms of resource sharing have been proposed to improve the network performance. From the perspective of any one user, when ignoring the interaction among users, all such schemes reduce to various forms of providing parallelism and, hence, increased throughput. In this paper, we argue that focusing on parallelism is by no means sufficient as it ignores the existence of many users with potentially similar strategies. In this paper, we illustrate the issue of resource sharing in the above context via a multi-queue multi-server problem. Although, our proposed model is not realistic in that it ignores the overhead inefficiencies, it does capture the trade-off between parallelism and reassembly/synchronization delay to a larger extent We use this model to provide analytical results in a special case of homogeneous users and servers. Furthermore, we prove the robustness of a certain locally optimal strategy to non-cooperation in a Nash equilibrium/strategy context.
展开▼
