Incorporating contribution-awareness into mesh-based Peer-to-Peer streaming systems

While Peer-to-Peer streaming has become increasingly popular over the Internet during recent years, the proper allocation of available resources among peers in a resource constraint environment, remains a challenging problem. In a resource constraint environment, the allocated resources and thus delivered quality to individual peers should be proportional to their contribution to the system, i.e., resource allocation should be contribution aware. This in turn results in fairness among peers and encourages active contribution from participating peers which is essential for scalability of P2P systems. However, contribution-aware resource allocation is challenging due to the distributed and dynamic nature of resources in P2P systems. In this paper, we present a tax-based contribution-aware scheme for live mesh-based P2P streaming approaches. In our proposed scheme, individual peers use a tax function to determine their number of parent peers (i.e., their share of resources) based on the number of their child peers (i.e., peers’ contributed resources) and the aggregate available resources in the system. We examine the behavior of a commonly used tax function, and describe how the contribution aware scheme can leverage the tax function. Through extensive simN. Magharei (B) · R. Rejaie Department of Computer and Information Science, University of Oregon, Eugene, OR 97403-1202, USA e-mail: [email protected] R. Rejaie e-mail: [email protected] Y. Guo Research and Innovation, Technicolor, 2 Independence Way, Princeton, NJ 08540, USA e-mail: [email protected] ulations we demonstrate the ability of our proposed scheme to properly allocate available resources among participating peers over a wide range of scenarios. We show that the amount of resources (i.e., bandwidth) is divided across peers proportional to their contribution and in our default simulation setting the median delivered quality to high bandwidth peers with high contribution is improved by 100%. We believe that our results shed an insightful light on the dynamics of resource utilization and allocation in the context of live mesh-based P2P streaming.