Auction-based P2P VoD streaming: Incentives and optimal scheduling

  • Authors:
  • Chuan Wu;Zongpeng Li;Xuanjia Qiu;Francis C. M. Lau

  • Affiliations:
  • The University of Hong Kong, Hong Kong;University of Calgary, Canada;The University of Hong Kong, Hong Kong;The University of Hong Kong, Hong Kong

  • Venue:
  • ACM Transactions on Multimedia Computing, Communications, and Applications (TOMCCAP) - Special Issue on P2P Streaming
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Real-world large-scale Peer-to-Peer (P2P) Video-on-Demand (VoD) streaming applications face more design challenges as compared to P2P live streaming, due to higher peer dynamics and less buffer overlap. The situation is further complicated when we consider the selfish nature of peers, who in general wish to download more and upload less, unless otherwise motivated. Taking a new perspective of distributed dynamic auctions, we design efficient P2P VoD streaming algorithms with simultaneous consideration of peer incentives and streaming optimality. In our solution, media block exchanges among peers are carried out through local auctions, in which budget-constrained peers bid for desired blocks from their neighbors, which in turn deliver blocks to the winning bidders and collect revenue. With strategic design of a discriminative second price auction with seller reservation, a supplying peer has full incentive to maximally contribute its bandwidth to increase its budget; requesting peers are also motivated to bid in such a way that optimal media block scheduling is achieved effectively in a fully decentralized fashion. Applying techniques from convex optimization and mechanism design, we prove (a) the incentive compatibility at the selling and buying peers, and (b) the optimality of the induced media block scheduling in terms of social welfare maximization. Large-scale empirical studies are conducted to investigate the behavior of the proposed auction mechanisms in dynamic P2P VoD systems based on real-world settings.