Network Processors are multicore processors capable of processing network packets at wire speeds of multi-Gbps. Due to their high performance and programmability, these processors have become the main computing elements in many demanding network processing equipments like enterprise, edge and core routers. With the ever increasing use of the internet, the processing demands of these routers have also increased. As a result, the number and complexity of the cores in network processors have also increased. Hence, efficiently managing these cores has become very challenging. This dissertation discusses two main issues related to efficient usage of large number of parallel cores in network processors: (1) How to allocate work to the processing cores to optimize performance? (2) How to meet the desired performance requirement power efficiently? This dissertation presents the design of a hash based scheduler to distribute packets to cores. The scheduler exploits multiple dimensions of locality to improve performance while minimizing out of order delivery of packets. This scheduler is designed to work seamlessly when the number of cores allocated to a service is changed. The design of a resource allocator is also presented which allocates different number of cores to services with changing traffic behavior. To improve the power efficiency, a traffic aware power management scheme is presented which exploits variations in traffic rates to save power. The results of simulation studies are presented to evaluate the proposals using real and synthetic network traces. These experiments show that the proposed packet scheduler can improve performance by as much as 40% by improving locality. It is also observed that traffic variations can be exploited to save significant power by turning off the unused cores or by running them at lower frequencies. Improving performance of the individual cores by careful scheduling also helps to reduce the power consumption because the same amount of work can now be done with fewer cores with improved performance. The proposals made in this dissertation show promising improvements over the previous work. Hashing based schedulers have very low overhead and are very suitable for data rates of 100 Gbps and even beyond. / text

Nous proposons une nouvelle structure de données distribuée et scalable appelée LH*RSP2P conçue pour les environnements pair à pair(P2P).Les données de l'application forment un fichier d’enregistrements identifiés par les clés primaires. Les enregistrements sont dans des cases mémoires sur des pairs, adressées par le hachage distribué (LH*). Des éclatements créent dynamiquement de nouvelles cases pour accommoder les insertions. L'accès par clé à un enregistrement comporte un seul renvoi au maximum. Le scan du fichier s’effectue au maximum en deux rounds. Ces résultats sont parmi les meilleurs à l'heure actuelle. Tout fichier LH*RSP2P est également protégé contre le Churn. Le calcul de parité protège toute indisponibilité jusqu’à k cases, où k ≥ 1 est un paramètre scalable. Un nouveau type de requêtes, qualifiées de sûres, protège également contre l’accès à toute case périmée. Nous prouvons les propriétés de notre SDDS formellement par une implémentation prototype et des expérimentations. LH*RSP2P apparaît utile aux applications Big Data, sur des RamClouds tout particulièrement / We propose a new scalable and distributed data structure termed LH*RSP2P designed for Peer-to-Peer environment (P2P). Application data forms a file of records identified by primary keys. Records are in buckets on peers, addressed by distributed linear hashing (LH*). Splits create new buckets dynamically, to accommodate inserts. Key access to a record uses at most one hop. Scan of the file proceeds in two rounds at most. These results are among best at present. An LH*RSP2P file is also protected against Churn. Parity calculation recovers from every unavailability of up to k≥1, k is a scalable parameter. A new type of queries, qualified as sure, protects also against access to any out-of-date bucket. We prove the properties of our SDDS formally, by a prototype implementation and experiments. LH*RSP2P appears useful for Big Data manipulations, over RamClouds especially.