Socially Augmented Content Distribution in Hybrid Networks

Internet traffic is experiencing a great growth whose end cannot be foreseen. Whether it will always be possible to upgrade the infrastructure fast enough to keep up with the increasing demand is uncertain. On any account this undertaking will be expensive and the consequences that arise from the thereby growing complexity are hard to predict. For many files delivered via Internet a large degree of redundancy can be observed. Since such transfers use unicast transmission mode, the induced traffic increases linearly with the number of downloads. This is a challenge for the delivery network used and for the servers. This thesis argues that the peaks in Internet traffic can be truncated and the aggregated amount of bytes transferred be significantly reduced by shifting this redundant traffic to a broadcast network. Thereby the unicast and broadcast network are combined by means of a hybrid network that allows dynamic routing of data via either of them. Based on this infrastructure a socially augmented content distribution model relying on P2P technology is defined that manages the network selection and the delivery of data. It aims at delivering files to user caches prior to the demand and thus can level out the fluctuations in Internet and server load. Thereby it makes use of social networks by means of capitalizing information about social ties, similarities and mutual influence among individuals. It is shown that by this approach the Internet traffic can be significantly reduced while only requiring a comparatively small amount of broadcast network capacity. At the same time the user perceived quality of service is increased by means of a reduced delivery time.