Modeling Priority-Based Incentive Policies for Peer-Assisted Content Delivery Systems

Abstract

Abstract. Content delivery providers can improve their service scalability and offload their servers by making use of content transfers among their clients. To provide peers with incentive to transfer data to other peers, protocols such as BitTorrent typically employ a tit-for-tat policy in which peers give upload preference to peers that provide the highest upload rate to them. However, the tit-for-tat policy does not provide any incentive for a peer to stay in the system beyond completion of its download. This paper presents a simple fixed-point analytic model of a priority-based incentive mechanism which provides peers with strong incentive to contribute upload bandwidth beyond their own download completion. Priority is obtained based on a peer's prior contribution to the system. Using a two-class model, we show that priority-based policies can significantly improve average download times, and that there exists a significant region of the parameter space in which both high-priority and low-priority peers experience improved performance compared to with the pure tit-for-tat approach. Our results are supported using event-based simulations.

Using Torrent Inflation to Efficiently Serve the Long Tail in Peer-Assisted Content Delivery Systems

“…Yang et al [7] propose a cross-torrent tit-fortat scheme in which unchoking is done based on the aggregate download rate from each candidate peer across all torrents (rather than just within a single torrent). Other prior work has proposed that inter-swarm exchanges be incentivized through propagation of peer reputation [11], incentive-based token schemes [12], and/or history-based priority rules [13]. These works are complimentary to ours, and could perhaps be used to incentivize peer participation in torrent inflation.…”

Caching-Based Multi-Swarm Collaboration for Improving Content Availability in BitTorrent

“…As the incentive, they propose a cross-swarm tit-for-tat scheme in which unchoking (i.e., uploading chunks to a peer who issues a content request) is done based on the aggregate download rate from each candidate peer across all swarms. Other several work also tries to make inter-swarm content exchange happens by proposing various incentive mechanisms such as incentive-based token scheme [7], history-based priority scheme [8], and peer reputation propagation [9].…”

On Zipf Models for Probabilistic Piece Selection in P2P Stored Media Streaming

“…Centralized approaches typically rely on the classic client-server architecture, augmented with scalable streaming protocols, IP multicast, or combinations thereof. Decentralized approaches include the peer-to-peer (P2P) paradigm [1], [7], [12], [15], [19], [29], [31], as well as peerassisted streaming [8], [9], [16]. Hybrid approaches using Content Delivery Networks (CDN) or cloud-based streaming have also emerged recently.…”