Framework Architecture for Decentralized Communications

Peer-to-Peer (P2P) systems represent a paradigm shift from the traditional client/server architecture. Over the past decade, P2P technologies have proven themselves as a viable option for providing services in the Internet. This success has resulted in initiatives to develop standards-based P2P protocols and services. One of the major initiatives in this area is Peer-to-Peer Session Initiation Protocol (P2PSIP), a suite of communication protocols that enable the Session Initiation Protocol (SIP) to decentralize its functions. P2PSIP is being standardized in the Internet Engineering Task Force (IETF). This dissertation presents a framework architecture for decentralized communications that is built around P2PSIP and the set of technologies it uses, including the REsource LOcation And Discovery (RELOAD) P2P signaling protocol, Chord Distributed Hash Table (DHT) algorithm, and the Interactive Connectivity Establishment (ICE) Network Address Translator (NAT) traversal solution. The framework presented in this dissertation is a set of reusable and modular software components that can be used in a flexible manner either individually or in different combinations to support the needs of a broad set of applications and use cases. Due to its flexibility and modularity, the framework is not an integrated architecture, tightly coupled set of components, or a purpose-built software platform whose components cannot function individually or are not interchangeable. The framework and all of its components were implemented as a part of the work on this dissertation. In the dissertation, the performance of the implementation of the framework and its components is evaluated using real-world prototypes and simulators. The focus is on evaluating the performance of DHT maintenance routines, ICE-based NAT traversal, the operations the framework provides to applications, and the performance of the implementation of the framework in mobile environments. Based on the performance analysis, missing features and performance bottlenecks are identified. The performance bottlenecks are addressed and the missing features are added by designing new components to complete the framework. These components include self-tuning, service discovery, M2M communication, and session setup delay optimization components. The contributions of this dissertation can be divided into three categories. First, the delays associated with using the services and operations provided by the framework are analyzed and optimized. Second, the overlay network platform that the framework provides is evaluated and extended. Finally, the framework is applied to new use cases. The overall result of the work is a framework architecture for decentralized communications that is scalable, adaptive, generic, modular, based on emerging standards, and has high performance.