Cross-Layer P2P Traffic Engineering in Content-based Networks

Peer-to-peer (P2P) content sharing applications account for a significant fraction of the traffic volumes and is expected to increase [1]. Data is distributed to a large population of end client peers from end source peers in P2P systems, without the need for big investments to deploy servers. The costs of the content distribution are thus shared among end users and Internet service providers (ISPs). Consequently, negative impacts, such as increased inter-ISP traffic in particular, have become critical issues that need to be kept low, due to that most popular P2P protocols are not designed to be aware of network topology. On the other hand, substantial access burden at source peers’ side can be increased due to their limited uploading bandwidth capacities, compared to the massive data demand. The top-level technical objectives of this work are thus to 1) achieve optimised usage of network resources through reducing P2P content traffic, and at the same time, 2) to provide enhanced network support to P2P applications. Specifically, we address the above issues in mainly two ways: First, in order to reduce the P2P traffic load especially over the costly inter-ISP links, we propose an advanced hybrid scheme for peer selections across multi-domains, by promoting cooperation and non-cooperation among adjacent ISPs. An analytical modelling framework is developed for analysing inter-domain peer selection schemes concerning ISP business policies. Our analytical modelling framework can be used as a guide for analysing and evaluating future network-aware P2P peer selection paradigms in general multi-domain scenarios. Second, with the concern of improving service quality for P2P users in terms of content access delay and transmission delay, we propose an intelligent in-network caching scheme enabled by Information-Centric Networking (ICN). A simple analytical modelling framework is developed to quantitatively evaluate the efficiency of the proposed in-network caching policy. We further design an ICN-driven protocol for the efficient P2P content delivery with in-network caching support. Bloom Filter (BF) techniques are adopted to save cache space and also to reduce communication overhead. A P2P-like content delivery simulator with In-network caching functionalities is built, with which extensive simulation experiments are conducted to validate the analytical results and to further prove the efficiency of the proposed caching scheme.