Abstract
Wide-area distributed systems, including grid computing, Peer-to-Peer (P2P) computing and wireless distributed computing have recently emerged as one of the most popular applications in today's Internet. Millions of users are connected to each other to share their limited resources, such as CPU cycles, storage spaces, spectrum resources and network bandwidths, and to deliver a large variety of contents in a distributed manner with low transmission cost. With the success of these widely deployed applications, how to design efficient, flexible and stable resource sharingstrategies by considering the characteristics of resources, such as scalability, heterogeneity, dynamics and locality becomes very important in order to better handle challenges brought by the enormous (and constantly increasing) amount of user nodes and contents in new applications. In this dissertation, our major objective is to clearly address these challenges particularly in the distributed P2P storage system, distributed P2P live streaming system and distributed cognitive radio network, and investigateresource sharing strategies to overcome the challenges for achieving the construction of a more efficient, flexible and stable wide-area distributed resource sharing environment.In this dissertation, we first study an overlay construction approach in P2P storage systems by exploiting the locality of peers in the underlying network. Specifically, in contrast with the P2P storage system only aware of data-locality, we propose a routing-locality-aware hybrid structured/unstructured P2P storage system to solve the routing inefficiency problem and to achieve scalable and reliable file distribution among peers. In the second part of the dissertation, we propose a two-layered hybrid tree/mesh overlay structure by exploiting the locality andflexibility that makes the overlay capable of adjusting itself according to the dynamics of peers and enhances the streaming performance of previous mesh-based P2P live streaming systems.Finally, we study a coalitional game approach to resource allocation in a multi-channel cooperative cognitive radio network (CCRN) with multiple primary users (PUs) and multiple secondary users (SUs). Since previous works only consider a single channel scenario that involves one PU and multiple SUs, this consideration presents a simplification for practical scenarios where there are typically multiplechannels and multiple PUs that coexist in the coverage area of a base station in the cellular system. Thus, we investigate the cooperation strategies between multiple PUsand multiple SUs in a multi-channel CCRN, and apply the solution concepts of the coalitional game, namely the core and the Shapley value, respectively, to characterize the stability and fairness of the payoff allocation from the aggregate utility among rational users. In this dissertation, we have shown that our proposed approaches ofresource sharing in various distributed systems fulfilled their design objectives, using both theoretical analysis and extensive simulations.