Print Email Facebook Twitter Communicating with low latency peers Title Communicating with low latency peers: Building a low latency overlay in P2P networks Author van IJzendoorn, Bas (TU Delft Electrical Engineering, Mathematics and Computer Science) Contributor Pouwelse, Johan (mentor) Rellermeyer, Jan (graduation committee) Spaan, Matthijs (graduation committee) Degree granting institution Delft University of Technology Date 2018-07-09 Abstract In decentralized applications requirements on latency are high. For instance, in tradingapplications a millisecond latency advantage can save companies millions of dollars. Inthis thesis a design is proposed for an overlay in a decentralized peer-to-peer systemthat lowers the latency between connected peers. The overlay is designed and imple-mented on top of Tribler, a BitTorrent client developed at Delft University of Technol-ogy. The client continually connects toward the peers with the lowest latency to keepan established connection towards them. Next to that, the introductions in the peerdiscovery mechanism are chosen in such a way that introduced peers have a low la-tency toward each other. To achieve this latency estimation algorithms are used thatestimate the latency between arbitrary peers in the network. These algorithms requirea limited amount of measured latencies between peers in the network and can estimatethe others. Latencies are continually measured toward peers that have an establishedconnection to a client. These measured latencies are shared to other peers for furtherimprovements on the quality of the latency estimation algorithms. The overlay design isevaluated on the accuracy of the latency estimation algorithms, computation time andto what extent the connections of a peer are low latency. The results show that the peersin the overlay are able to connect toward their lowest latency peers despite low accuracyof the latency estimation algorithms. Subject latencyoverlaydistributed systems To reference this document use: http://resolver.tudelft.nl/uuid:3bf51816-d0e3-48fe-a794-6a4f5713baae Part of collection Student theses Document type master thesis Rights © 2018 Bas van IJzendoorn Files PDF final_version.pdf 1.81 MB Close viewer /islandora/object/uuid:3bf51816-d0e3-48fe-a794-6a4f5713baae/datastream/OBJ/view