70 likes | 160 Views
P2P Streaming Protocol Pro-incentive Parameters draft-zeng-ppsp-protocol-pro-incentive-para-01 IETF79 Meeting Wenjun (Kevin) Zeng & Yingjie Gu Huawei Technologies. Introduction. Lack of cooperation is one of the key problems that confront today's P2P systems. Free riding Starvation
E N D
P2P Streaming Protocol Pro-incentive Parametersdraft-zeng-ppsp-protocol-pro-incentive-para-01 IETF79 Meeting Wenjun (Kevin) Zeng & Yingjie GuHuawei Technologies
Introduction • Lack of cooperation is one of the key problems that confront today's P2P systems. • Free riding • Starvation • This proposal intends to • analyze various incentive mechanisms/dishonest manipulation • extract the common pro-incentive parameters, and • propose to incorporate these pro-incentive parameters in information exchanges in the P2P streaming protocols. • NOT intend to specify a particular incentive mechanism in the P2P streaming protocol.
Reward based on Peer Contribution • A typical metric for peer contribution: • The amount of upload a peer has contributed. • E.g., BitTorrent uses a bilateral mechanism: tit-for-tat. • Contribution-based approach is vulnerable to strategic manipulation by peers. • E.g., the Discount Parameter (DP) attack is an incentive threat that exploits the tit-for-tat based unchoking. • DP: Determines the degree to which the payoff of each move is discounted relative to the previous move • If the DP is small, players might defect and not worry about future consequences • In p2p systems, Piece Rarity is a DP
Discount Parameter Exploitation Alice Bob Connects to Bob, Request for download piece X Determines whether Alice is a good uploader ?? ALICE UNCHOKED Ok send X.. Request for rarest piece Alice has: Y Alice : “Y is a rare piece. None of the pieces Bob has is rare enough compared to X” BOB CHOKED Bob loses a round of download Alice starts looking for other peers
Starvation • All peers have all pieces, except a few. • Origins: • Choices made by peers in piece and peer selection • Positive Discrimination • Use of rarest first approach • Can positive discrimination guarantee safe state? • Peers benefit from under-reporting, increasing demand for pieces they have and faster completion. • Why do we have Under Reporting Peers (URPs): • The orthogonal treatment of Piece Rarity and Peer Selection
DP/Starvation Remedy • Peer selection not only depends on the uploading bandwidth of the candidate peers, but also how valuable the pieces they have uploaded are. • Piece Rarity parameter • local availability (in the target swarm) of a piece: • a rarer downloaded piece has more value to the swarm • # of upload slots a candidate peer has: • long term benefit can be expected from peer with more uploading potential • completion factor (i.e., ratio of # of pieces that a peer has to total #of pieces of that file) of the candidate peer: • peer with high completion factor is a good one to maintain a good upload/download relation with. • the contention in the swarm (i.e., the ratio of total # of peers to total # of seeds): • high contention implies more strategic value of a piece.
Pro-incentive Protocol Parameters • Important for peers to collect pro-incentive parameters to support incentive mechanisms and propose to include the parameters in Tracker and Peer Protocols. • no_upload_slots: a peer's upload bandwidth (i.e., number of upload slots a peer has). • bytes_uploaded: total amount of data that a peer has uploaded • bytes_downloaded: total amount of data that has been downloaded from a peer • chunk_nos: total # of chunks of a file that a peer has. • seed_nos: total # of seeds. • peer_nos: total # of peers. • chunk_copies_swarm: chunk availability, i.e., total # of copies of a chunk available in the swarm.