2006-2007 Seminar Series

Scalable on-demand content delivery systems, designed to effectively handle increasing request rates, typically use service aggregation or content replication techniques. Service aggregation relies on one-to-many communication techniques, such as multicast, to efficiently deliver content from a single sender to multiple receivers. With replication, multiple geographically distributed replicas of the service or content share the load of processing client requests and enable delivery from a nearby server.

This talk concerns the scalability and performance of download protocols. Three different contexts are considered. The first is that of a single server employing aggregation to serve large numbers of concurrent clients. Both existing and new protocols are evaluated against analytic lower bounds developed in this work.

The second context considered is that of systems exploiting both service aggregation and replication. Replica selection policies must take into consideration the basic tradeoff between locality of service (maximized by selecting the nearest replica), and efficiency of use of server resources (maximized by selecting the replica at which service can be shared among the largest number of clients). Classes of policies of differing complexities are compared within the context of a simple cost model.

The final context considered is that of a streaming (rather than download) service, in which playback can begin well before the entire media file is received. An approach is proposed wherein variants of peer-assisted scalable download protocols are used to provide such a service.

Niklas Carlsson is currently a Ph.D. candidate in the Department of Computer Science at the University of Saskatchewan, Saskatoon, Canada. He received his M.Sc. degree in engineering physics from Umeå University, Umeå, Sweden, in 2000. His research interests are in the areas of performance evaluation of distributed systems and computer networks. They encompass content delivery systems, peer-to-peer systems, wireless sensor networks, as well as modeling and characterization of the Internet.