System designs for bulk and user-generated content delivery in the internet

This thesis proposes and evaluates new system designs to support two emerging Internet workloads: (a) bulk content, such as downloads of large media and scientific libraries, and (b) user-generated content (UGC), such as photos and videos that users share online, typically on online social networks (OSNs). Bulk content accounts for a large and growing fraction of today’s Internet traffic. Due to the high cost of bandwidth, delivering bulk content in the Internet is expensive. To reduce the cost of bulk transfers, I proposed traffic shaping and scheduling designs that exploit the delay-tolerant nature of bulk transfers to allow ISPs to deliver bulk content opportunistically. I evaluated my proposals through software prototypes and simulations driven by real-world traces from commercial and academic ISPs and found that they result in considerable reductions in transit costs or increased link utilization. The amount of user-generated content (UGC) that people share online has been rapidly growing in the past few years. Most users share UGC using online social networking websites (OSNs), which can impose arbitrary terms of use, privacy policies, and limitations on the content shared on their websites. To solve this problem, I evaluated the feasibility of a system that allows users to share UGC directly from the home, thus enabling them to regain control of the content that they share online. Using data from popular OSN websites and a testbed deployed in 10 households, I showed that current trends bode well for the delivery of personal UGC from users’ homes. I also designed and deployed Stratus, a prototype system that uses home gateways to share UGC directly from the home.