Distribution of Multimedia Streams to Mobile Internet Users

In this thesis, we consider the efficient distribution of live and scheduled multimedia content (e.g., radio and TV broadcasts) to mobile users via a ubiquitous wireless Internet. The objective is to design and develop a content delivery system that (1) enables content owners to deliver their multimedia content to a large number of heterogeneous receivers, and (2) enables receivers to continuously receive that content, independent of their location and the network they connect to. Previous investigations into this topic have shown that multimedia content can be efficiently distributed through an overlay network that consists of multiple distributed proxy servers. In this thesis, we extend this concept to the distribution of live and scheduled multimedia content through multiple aggregators. An aggregator is an intermediary content provider that aggregates live multimedia content from various content sources (e.g., news services) and delivers it to mobile users through a pool of proxy servers. The availability of the same content through multiple aggregators enables mobile users to switch from one aggregator to another, thus alternately receiving the same content from different aggregators. The service area of an aggregator may be restricted to a certain set of networks, in which case switching to such an aggregator also requires mobile hosts to handoff to a network that is part of the aggregator’s service area. We call the system that switches a mobile host to another aggregator the ALIVE system, which stands for Aggregator Switching System for Mobile Receivers of Live Multimedia Streams. We concentrate on the ‘front-end’ of the ALIVE system, which supports mobile users and mobile hosts, aggregators, and wireless networks. We particularly focus on the signaling interactions between mobile hosts and aggregators and do not consider the details of the multimedia content itself. The design of the ALIVE system is based on the ALIVE business network, which is a network of business roles (consisting of roles such as ‘aggregator’ and ‘end-user’) that describes the relations that may exist between domains involved in the distribution of live multimedia content through multiple aggregators. In line with current trends in content distribution, the ALIVE business network consists of an application-level part (an overlay that consists of content sources and aggregators) and a network-level part (consists of providers of basic Internet access). We capture the properties of the relations in the business network in so-called "agreements". The ALIVE business network uses the notion of a channel to refer to a particular piece of content (e.g., a TV broadcast). To further increase the number of potential receivers, aggregators in the ALIVE business network are able to transmit channels in various configurations. A configuration delivers a channel in a specific perceptual quality and requires a welldefined level of resources (e.g., network bandwidth). Aggregators may choose to support a relatively small number of configurations, thus striking a balance between per-user personalization of a channel (e.g., by delivering a channel in a configuration that is tailored to the instantaneous bandwidth available to a specific user’s mobile host) and no personalization at all (i.e., everybody receiving a channel in the same configuration). To support roaming users, aggregators establish application-level roaming agreements between each other. These agreements enable users to receive channels from multiple aggregators (e.g., at different locations) while having a subscription with only a few of them (typically one). Application-level roaming agreements define the configurations in which a user can receive channels from an aggregator with which the user does not have a subscription (called a foreign aggregator). An aggregator may be bound to a specific set of networks through a socalled "binding agreement" with Internet access providers. In the ALIVE business network, such an aggregator is called a local aggregator because the binding agreement restricts its service area to the networks of the involved access provider. An Internet access provider may use local aggregators to offer exclusive channels or channel configurations to the users that connect to the access provider’s networks (cf. the walled-garden models that contemporary cellular operators typically use). The ALIVE system itself enables mobile users to roam in an unrestricted manner while continuously receiving a channel. The system transparently switches mobile hosts from one aggregator to another and executes handoffs on the mobile host’s network interfaces. The system switches a mobile host to the aggregator that provides a certain channel in the best configuration, where ‘best’ is defined by the preferences of the end-user. This makes the ALIVE system a user-centric system. The ALIVE system is scalable because most of its logic resides on mobile hosts (mobile-controlled switching). The system’s operation is policycontrolled, which enables stakeholders in the ALIVE business network to flexibly change the rules that the ALIVE system uses to make decisions (e.g., when to look for another available configuration of a channel). The ALIVE system contains an application-level protocol, which we realized using the Session Initiation Protocol (SIP) and the Session Description Protocol (SDP), both of which are Internet standards. We deployed our implementation in a small-scale testbed with different types of networks, which represent the ‘beyond 3G’ Internet environment in which the ALIVE protocol is intended to operate. Trough an analysis of our SIP-based implementation of the ALIVE protocol, we obtained quantitative information on how to smoothly execute switches between aggregators. Our analysis concentrates on the extra delay introduced by the ALIVE protocol in a contemporary wireless Internet environment, specifically consisting of 802.11 hotspots and UMTS/GPRS overlays. We focus on the operation of the ALIVE protocol immediately after a handoff to another 802.11 access provider, which is where the ALIVE protocol typically comes into play. After a handoff, the ALIVE system usually first attempts to discover the configurations in which the user can receive a channel from the local aggregators on the new network. At the edges of 802.11 cells, this may result is a significant delay because of the exponential back-off mechanism that SIP uses to recover from packet loss. Our analysis consists of two parts: (1) a heuristic analysis of the application and network-level delay components involved in a typical switch and an estimation of their best-case values, and (2) an empirical analysis of the delay introduced by SIP transactions under various 802.11 network conditions. The analysis shows that the ALIVE system usually experiences little delay, except at the very edge of an 802.11 cell. Based on our implementation and measurement work, we conclude that the ALIVE system is a feasible system that provides a clear contribution to the multimedia-everywhere vision