Social media content distribution: measurement and enhancement

Author: 
Date created: 
2012-06-08
Identifier: 
etd7234
Keywords: 
Social media
Video sharing
Social networking
Measurement
Enhancement
Abstract: 

In the past decade, such popular social media as YouTube, Facebook, and Twitter have substantially changed the content distribution landscape and become an important part in people's everyday life. Extensive research works have been conducted to understand them in the recent years. However, a number of new features emerge and a number of directions are yet to be explored. This thesis largely extends the current research efforts on social media content distribution by measurements and enhancements. We first analyze YouTube Insight dataset from a partner's view, revealing the inherent relationship among various metrics which affect the popularity of the videos. Our findings facilitate YouTube partners to adapt their content deployment and user engagement strategies to generate more views and subsequently increasing their revenues. We also take an important step towards understanding the characteristics of video spreading in social media, examining the user behaviour and the spreading structure. We propose an epidemic model to capture the process of video spreading, which serves as a valuable tool for workload synthesis, traffic prediction, and resource provisioning. Motivated by our measurement and a user questionnaire survey, we reveal a new scenario of coexistence of sharing and streaming. We propose a novel system that leverages stable storage users and yet inherently prioritizes living streaming flows, providing better scalability, robustness, and streaming quality. On the other hand, the recently emerged cloud service is a promising solution to the huge demands of bandwidth and storage from the growing social media. However the existing works on partitioning social media contents only focus on preserving the social relationship. We take an important factor, user access pattern, into account, and formulate the problem as a constrained k-medoids clustering problem. Our solution shows significant decrease of the access deviation and flexible preservation of the social relationship.

Document type: 
Thesis
Rights: 
Copyright remains with the author. The author granted permission for the file to be printed and for the text to be copied and pasted.
File(s): 
Supervisor(s): 
Jiangchuan Liu
Department: 
Applied Science: School of Computing Science
Thesis type: 
(Thesis/Dissertation) Ph.D.
Statistics: