Quality-consciousness in Large-scale Content Distribution in the Internet

Content distribution is the primaryfunction of the Internet today.Technologies like multicast andpeer-to-peer networks hold the potentialto serve content to large populations ina scalable manner. While multicastprovides an efficient transportmechanism for one-to-many andmany-to-many delivery of data in anInternet environment, the peer-to-peernetworks allow scalable content locationand retrieval among large groups ofusers in the Internet.Incorporating quality-consciousness inthese technologies is necessary toenhance the overall experience ofclients. This dissertation focuses onthe architectures and mechanisms toenhance multicast and peer-to-peercontent distribution throughquality-consciousness. In particular,the following aspects ofquality-consciousness are addressed: 1)client latency, 2) servicedifferentiation, and 3) content quality.Data analysis shows that the existingmulticast scheduling algorithms behaveunfairly when the access conditions forthe popular files changes. They favorthe popular files while penalizing thefiles whose access conditions have notchanged. To maintain the client latencyfor all files under dynamic accessconditions we develop a novel multicastscheduling algorithm that requires nochange in server provisioning.Service differentiation is a desirablefunctionality for both multicast andpeer-to-peer networks. For multicast,we design a scalable and low overheadservice differentiation architecture.For peer-to-peer networks, we focus on aprotocol to provide different levels ofservice to peers based on theircontributions in the system.The ability to associate reliablereputations with peers in a peer-to-peernetwork is a useful feature of thesenetworks. Reliable reputations can helpestablish trust in these networks andhence improve content quality. They canalso be used as a substrate for aservice differentiation scheme for thesenetworks. This dissertation developstwo methods of tracking peer reputationswith varying degrees of reliability andoverheads.