Le besoin des applications de streaming multimédia en temps réel sur Internet ne cesse d’accroître. Dans ce contexte, les réseaux Pair-à-Pair (P2P) jouent un rôle prépondérant afin d’assurer une transmission robuste et extensible (scalable) de contenu multimédia à grand-échelle. Cependant, le déploiement des applications de streaming multimédia à travers les réseaux P2P présente plusieurs verrous. En effet, les applications de streaming vidéo nécessitent une garantie de performances temps réel en termes de : délai minimal de bout-en-bout, faible taux de perte de paquets et garantie de la bande passante. Par ailleurs, un réseau P2P est composé d’un ensemble de terminaux interconnectés à travers des réseaux hétérogènes, n’ayant pas forcément les mêmes caractéristiques requises afin d’offrir la même qualité vidéo aux clients finaux.Dans cette thèse, nous analysons différents problèmes liés au streaming en-direct (Live) ou à la demande (VoD) de la vidéo sur les réseaux P2P, et nous proposons un ensemble de mécanismes pour y remédier. / There is an increasing demand for efficient deployment of real-time multimedia streaming applications over Internet. In this context, Peer-to-Peer (P2P) networks are playing an important role for supporting robust and scalable transmission of multimedia content to large-scale and massive receivers. P2P networks consist of different heterogeneous networks and devices, which may not have symmetric characteristics to offer the same video quality to end clients. The deployment of streaming applications such as live and on-demand over P2P networks is challenging. Indeed, video streaming applications require real-time performance guarantee in terms of low end-to-end delay, low packet loss and guaranteed available bandwidth. Moreover, P2P video streaming also experiences the problems of long start-time and churn-induced instability. The situation leads to affect the viewing experience of the users in P2P networks.The aim of this thesis is to analyze these issues and to propose quality-aware mechanisms for both live and on-demand streaming over P2P networks. Our contributions in this dissertation are threefold. First, we propose a small world overlay organization mechanism allowing efficient organization of peers to reduce the end-to-end delay for live streaming applications. Second, we propose a mechanism for cooperative prefetching with differentiated chunks scheduling suitable for P2P video-on-demand (VoD) to efficiently handle seek operations while reducing the response latency and increasing the hit ratio. Finally, we present a smoothing mechanism for layered streaming in P2P networks. The mechanism aims to reduce the number of layer changes under varying network conditions while at the same time achieving a high delivery ratio.

Data Prefetching is a well-known technique to speed up applications wherein hardware prefetchers or compilers speculatively prefetch data into caches closer to the processor to ensure it’s readily available when the processor demands it. Since incorrect speculation leads to prefetching useless data which, in turn, results in wasting memory bandwidth and polluting caches, prefetch mechanisms are usually conservative and prefetch on spotting fairly regular access patterns only. This gives the programmer with a knowledge of application, an opportunity to insert fine-grain software prefetches in the code to clinically prefetch the data that is certain to be demanded but whose access pattern is not too obvious for hardware prefetchers or compiler to detect. In this study, the author demonstrates the performance improvement obtained by such programmer-inserted prefetches with the case study of an FMM (Fast Multipole Method) VList application kernel run with several different configurations. The VList computation requires computing the Hadamard product of matrices. However, the way each node of the octree is stored in the memory, leads to indirect accessing of elements where memory accesses themselves are not sequential but the pointers pointing to those memory locations are still stored sequentially. Since compilers do not insert prefetches for indirect accesses, and to hardware, the access pattern appears random, programmer-inserted prefetching is the only solution for such a case. The author demonstrates the performance gain obtained by employing different prefetching choices in terms of what all structures in the code to prefetch and which level of cache to prefetch those to and also presents an analysis of the impact of different configuration parameters on performance gain. The author shows that there are several prefetching combinations which always bring performance gain without ever hurting the performance, and also identifies prefetching to L1 cache and prefetching all data structures in question, as the best prefetching recommendation for this application kernel. It is shown that this one combination gets the highest performance gain for most run configurations and an average performance gain of 10.14% across all run configurations. / text