Electronic Theses and Dissertations

Date of Award

2014

Document Type

Thesis

Degree Name

M.S. in Engineering Science

Department

Computer and Information Science

First Advisor

Feng Wang

Second Advisor

John N. Daigle

Third Advisor

Byunghyun Jang

Relational Format

dissertation/thesis

Abstract

BitTorrent is the most popular Peer-to-Peer (P2P) file sharing system widely used for distributing large files over the Internet. It has attracted extensive attentions from both network operators and researchers for investigating its deployment and performance. For example, recent studies have shown that under steady state, its rarest first scheme with the tit-for-tat mechanism can work very effectively and make BitTorrent near optimal for the generic file downloading process. However, in practice, the highly dynamic network environment, especially the notorious user churns prevalently existing in most peer-to-peer systems, can severely degrade the downloading performance. In this thesis, we first study on the limitations of BitTorrent under dynamic network environments, focusing on two scenarios where with our preliminary modeling and analysis, we clearly identify how network dynamics and peer churns can significantly degrade the performance. With these findings, we further propose a novel protocol named RaptorQP2P, which is based on RaptorQ coding, to overcome the limitations of current BitTorrent design and maximize the performance of P2P file distribution. The new protocol features two levels of RaptorQ encoding. At the top layer, the entire file is RaptorQ encoded to yield a collection of source blocks and repair blocks, and then each source and repair block is RaptorQ encoded independently to yield a collection of source symbols and repair symbols for the block. The symbols are independently transferred among the peers and when a sufficient number of distinct symbols for a particular block have been received, whether source or repair, the block can be reconstructed. The file can be reconstructed using a sufficient arbitrary number of distinct blocks. Our results show that RaptorQP2P can well handle the network dynamics as well as peer churns and significantly shorten the downloading completion time by up to 41.4% with excellent scalability on both file size and user population.

Concentration/Emphasis

Emphasis: Computer Science

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.