<p>In recent years, with the fast development of resource capability of both the Internet and personal computers, multimedia applications like video-on-demand (VOD) streaming have gained dramatic growth and been shown to be potential killer applications in the current and next-generation Internet. Scalable deployment of these applications has become a hot problem area due to the potentially high server and network bandwidth required in these systems.</p><p>The conventional approach in a VOD streaming system dedicates a media stream for each client request, which is not scalable in a wide-area delivery system serving potentially very large numbers of clients. Recently, various efficient delivery techniques have been proposed to improve the scalability of VOD delivery systems. One approach is to use a scalable delivery protocol based on multicast, such as periodic broadcast or stream merging. These protocols have been mostly developed for single-server based systems and attempt to have each media stream serve as many clients as possible, so as to minimize the required server and network bandwidth. However, the performance improvements possible with techniques that deliver all streams from a single server are limited, especially regarding the required network bandwidth. Another approach is based on proxy caching and content replication, such as in content delivery networks (CDN). Although this approach is able to effectively distribute load across multiple CDN servers, the cost of this approach may be high.</p><p>With the focus on further improving the system efficiency regarding the server and network bandwidth requirement, a new scalable streaming protocol is developed in this work. It adapts a previously proposed technique called hierarchical multicast stream merging (HMSM) to use a peer-to-peer delivery approach. To be more efficient in media delivery, the conventional early merging policy associated with HMSM is extended to be compatible with the peer-to-peer environment, and various peer selection policies are designed for initiation of media streams. The impact of limited peer resource capability is also studied in this work. In the performance study, a number of simulation experiments are conducted to evaluate the performance of the new protocol and various design policies, and promising results are reported.

¡@¡@Kademlia[1] provides a concept of using XOR between ID to present logical distance. As Kademlia uses logical distance to search, it cannot take physical distance into consideration, which can lead the result to be the closest in logical, but the farthest in physical. ¡@¡@In this paper, ping rate is used to represent the meaning of physical distance, and construct a physical network in Kademlia. The operations like lookup and search will take logical and physical distance into consideration at the same time, called PRKad. ¡@¡@The simulation result shows that PRKad can actually automatically choose the more efficiency nodes, which means the close nodes in physical network, to decrease the time using of the following operations.

Recently there has been an increasing demand for efficient mechanisms of carrying out computations that exhibit coarse grained parallelism. Examples of this class of problems include simulations involving Monte Carlo methods, computations where numerous, similar but independent, tasks are performed to solve a large problem or any solution which relies on ensemble averages where a simulation is run under a variety of initial conditions which are then combined to form the result. With the ever increasing complexity of such applications, large amounts of computational power are required over a long period of time. Economic constraints entail deploying specialized hardware to satisfy this ever increasing computing power. We address this issue in Dispatch, a peer-to-peer framework for sharing computational power. In contrast to grid computing and other institution-based CPU sharing systems, Dispatch targets an open environment, one that is accessible to all the users and does not require any sort of membership or accounts, i.e. any machine connected to the Internet can be the part of framework. Dispatch allows dynamic and decentralized organization of these computational resources. It empowers users to utilize heterogeneous computational resources spread across geographic and administrative boundaries to run their tasks in parallel. As a first step, we address a number of challenging issues involved in designing such distributed systems. Some of these issues are forming a decentralized and scalable network of computational resources, finding sufficient number of idle CPUs in the network for participants, allocating simulation tasks in an optimal manner so as to reduce the computation time, allowing new participants to join the system and run their task irrespective of their geographical location and facilitating users to interact with their tasks (pausing, resuming, stopping) in real time and implementing security features for preventing malicious users from compromising the network and remote machines. As a second step, we evaluate the performance of Dispatch on a large-scale network consisting of 10−130 machines. For one particular simulation, we were able to achieve up to 1500 million iterations per second as compared to 10 million iterations per second on one machine. We also test Dispatch over a wide-area network where it is deployed on machines that are geographically apart and belong to different domains.

There are two main architectures in VOIP system at present. First is peer-to-peer, it has highly scalable, fault-tolerant and also can lighten the number and reliance of server. But there is no standard protocol between peers with different architecture, cause the clients unable to communication with each other. This problem can be solved by communicating from one P2P network to another. Second is Client-Server, it has mass of research data, lots of actual products, and standard protocol. This architecture is the most perfect one with simple structure, easy to maintain, lower response time than peer-to-peer structure, and has a variety of additional services, for instance Voice Mail, conference call, etc. All the client need is to obey sip standard protocol and it can register to any sip proxy to make a phone call. The disadvantage is no server no use. These two architectures have both good side and bad side, none of them is absolutely perfect. Our thesis is proposed a all new idea about ¡§Hybrid¡¨, this idea combine P2P and Client-Server architecture together to design a flexible soft phone that can be used is normal condition to register to a proxy, or setup a P2P network instantly in our own local area network. Finally our DCHS Mechanism is workable even when the sip proxy is maintaining or failure the client can use this mechanism to call any other user outside the P2P network by sharing other peers¡¦ call history.

Resource location is a necessary operation for computer applications. In large scale peer-to-peer systems, random search is a scalable approach for locating dynamic resources. Current peer-to-peer systems can be partitioned into those which rely upon the Internet for message routing and those which utilize an overlay network. These two approaches result in different connectivity topologies. This thesis analyzes the effect of topological differences on the effectiveness of random search. After demonstrating the benefits of an overlay network, we propose a hybrid approach for resource location. Our proposed protocol provides deterministic searching capabilities which can help prevent request failures for sensitive applications.

The use of peer-to-peer applications is growing dramatically, particularly for sharing large files and software. Currently, peer-to-peer file sharing systems are playing a dominant role in the content distribution over Internet. Therefore understanding the impact of peer-to-peer traffic on networks is significant and some reaction is necessary in order to keep the Internet functioning efficiently. In this Thesis, we designed a P2P network flow management system, combining Classifier and Linux TC (Traffic Control). We analyzed the network traffic by using Classifier and classified the network traffic into P2P and NP2P (non-P2P). Linux TC is a tool for implementing the management policy. Finally, the management has implemented in real network environment and improved the network performance.

Instead of using the public switched telephone network, VoIP (Voice-over-IP) services exchange voice information over Internet. Therefore, VoIP can services with their advantage of low rates. SIP (Session Initiation Protocol) is scalable, easy to implement, and requires less setup time than its predecessor protocols. Therefore, SIP becomes one of the major signaling protocols used in VoIP. SIP uses Client-Server architecture which is simple and easy to maintain. But Client-Server architecture may cause problems while the clients increase. For the reason, IETF (The Internet Engineering Task Force) drafts discusses using P2P (Peer-to-Peer) architecture in SIP protocol to avoid the weaknesses of Client-Server architecture, and hopes to provide good voice quality by using P2P architecture. In this paper, we discuss how to design and implement an embedded P2PSIP(Peer-to-Peer Session Initiation Protocol) user agent based on SIP standard. First, we analyze all kind of the major P2P structures, and discuss the advantage and disadvantage of them. Second, our design needs to consider the embedded systems which have limited resources. Using cache is a good solution. Finally, we confront the problem and try to solve it.

published_or_final_version / Electrical and Electronic Engineering / Master / Master of Philosophy

published_or_final_version / Computer Science / Master / Master of Philosophy

The persistence of cooperation is a longstanding problem in the social and biological sciences. Recent advances of peer-to-peer (P2P) networks manifest as a promising platform to experiment and contribute to theories and algorithms on cooperation. In this thesis, by and large, we view P2P systems as an economy in which incentives are critical to stimulate contribution. Indeed, a P2P system can be considered as a society where different behaviors can emerge, and an empirical platform to understand cooperation and mimic evolving population. Specifically, we consider the problem of cooperation from two perspectives. First and foremost, autonomous nodes are strategic and selfish, who are reluctant to cooperate solely for public good. We investigate incentive scheme design for cooperation in P2P live media streaming networks. The general approach of protocol decomposition shows that practical incentives can only be guaranteed by efficient peer selection, due to stringent playback deadlines. Striker strategy is then proposed so as to align the optimal peer selection of heterogeneous nodes with social welfare maximization, the efficiency of which is validated by repeated game modeling and extensive simulations. The hidden philosophy is to coerce non-cooperative peers into cooperation by collectively implementing punishment threats. This is analogous to strikes and coercion implemented by organizations like unions in human society. On the other hand, just as node selfishness, competition and struggle for survival raise another problem for cooperation. Similar to human society and biological systems, we envision that diverse strategies—some are more exploitative, while others more altruistic—could be deployed by selfish participants to compete against interacting nodes and gain performance advantages. In such a variegated environment, our coevolutionary perspective aims to understand cooperation and rationalize the coexistence and success of diverse behaviors. Population games and evolutionary game theory provide analytical tractability, while learning and evolutionary dynamics are proposed to evolve strategies. / published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy