End-to-end Measurement-assisted Relay Selection In Peer-to-peer Networks

With the rapid development of the Internet applications and services,the end-to-end network architecture has been greatly reshaped by variant network terminals and network access modes.Peer-to-Peer(P2P)networks utilize the application-level control protocols to link the end-hosts for constructing a virtual network structure,namely overlay networks.P2P networking offers a new network architecture in data sharing,content delivery and streaming service.End-to-End measurements can provide application-specific requirements,measure-ments and evaluations for P2P networks.The application-layer delay measurement and the relay discovery and selection in P2P networks with the measurement assistance have been investigated in this dissertation for reliable and robust content distributions in P2P networks.The major contributions of this dissertation are listed as follows:1.This dissertation proposes to measure the application-layer delay in a browser-scripting network measurement platform.A delay measurement study using engineered scripts has been conducted in the proposed browser-based measurement platform.The in-vestigation contains a comparison study of three browser-scripting delay measurement methods,including Java applet,JSP and Flash ActionScript.The developed browser-based delay measurement testbed over the Internet allows different delay measurement tools to be evaluated in the same real network environment including typical Internet paths and the Baidu cloud.The decomposition of the end-to-end delay process in the above measurements reveals the difference and relationship between network-layer de-lay and application-layer delay.The measurement results characterize the stochastic properties of the application-layer delay over real Internet paths,and how these prop-erties varied from the underlying network layer delay.2.A measurement-assisted multi-path overlay routing algorithm in a stochastic approx-imation approach has been proposed for P2P networks.This proposed overlay rout-ing algorithm,called the dynamic stochastic shortest path(DSSP),is constructed in a stochastic programming framework by leveraging the actual delay of local links and the statistical delay distributions of non-local overlay links.DSSP requires the most recent delay measurement and the historical delay samples to improve the accuracy in finding the optimized relay paths.DSSP is fully distributed and requires only accurate local information.Simulation results show that the proposed stochastic single/multi-path routing algorithm is more resilient to measurement errors and achieved much lower packet delays by 20%-30%than the deterministic shortest path algorithms,which utilize the average link delays.Trace-driven simulation demonstrates the effectiveness of the path diversity provided by the proposed algorithm in reducing packet loss up to nearly 50%in simulations.3.The location-aware relay discovery and selection problem has been studied for large-scale P2P streaming networks.The network location can be measured directly or in-directly with the tradeoffs between timeliness,overhead and accuracy.The proposed two-phase approach achieves efficient relay discovery and accurate relay selection.In-direct measurements are used to narrow down a small number of high-quality relay candidates and the final relay selection is refined based on direct probing.This two-phase approach enjoys an efficient implementation using the Distributed-Hash-Table(DHT).The relay discovery is achieved efficiently utilizing the DHT-based search.This dissertation analyzed end-to-end measurements in details,and with the application-specific requirements of relay issues in P2P networks.An end-to-end delay measurement framework and corresponding relay selection algorithms have been designed and imple-mented,to improve the application-layer routing performance and optimize the network ar-chitecture.