Distributed storage systems provide fast and reliable access to data by intro- ducing redundancy for stored les. The most common approach of adding re- dundant information is by repetition and erasure codes. Two main processes in a distributed storage system are reconstruction of original le and regenerat- ing a new node. These two processes require bandwidth, which are termed as reconstruction-bandwidth and repair-bandwidth. The current literature treat- s these two processes separately. That is, there are methods to reduce the reconstructing bandwidth without considering the repair-bandwidth, and also there are methods to reduce the repair-bandwidth. We study these two pro- cesses together and try to jointly minimize the reconstruction-bandwidth and repair-bandwidth. We observe a method that has the minimum reconstruction- bandwidth might have large amount of repair-bandwidth and vice versa. We propose codes which minimizes the sum of the repair-bandwidth and reconstruction- bandwidth. The main contribution of this thesis is nding an value of repair locality r (number of nodes connected during node repair) and devising two coding meth- ods in which total bandwidth approximates to be half reduced compared with MSR and MBR when k ! 1 under the condition that 1 6 r 6 k.

Hamid Ghasemirahni, et al. have shown that the order of the network packets arrival at a datacenter, equipped with commodity servers, plays a significant role in the latency of processing these packets. The larger the burst of packets that are part of a flow and hence will be processed using the same instructions and data, the greater the utilization of the system’s caches and subsequently the lower the latency of their processing. However, there are many reason for the packets to not arrive in a burst. One of the main reason that is examined in this project is packet interleaving that takes place in the routers and switches along the path that the packets take from a computer to/from the datacenter. This project take a more general look at traffic arriving at a server via a tree of network devices on the uplink path to the server. The focus is to use scripts to create a tree of switches and conduct experiments with them by scripting Mininet. From these experiments we learned that the different algorithms that control the output queues of the network nodes across the network path play a significant role in packet interleaving. Furtheromore, experiments on Mininet host connectivity explained how with the Openflow protocol, Mininet controller set up rules in the switches of the network topology. Finally, experiments in TCP throughput showed the limiting factors of a TCP connection between the server and a host while many provided traffic flows illustrate common behaviors of packet interleaving that occurs due to the switches. / Hamid Ghasemirahni, et al. har visat att ordningen på nätverkspaketens ankomst till ett datacenter, utrustat med varuservrar, spelar en betydande roll i fördröjningen av bearbetningen av dessa paket. Ju större burst av paket som är en del av ett flöde och därför kommer att behandlas med samma instruktioner och data, desto större utnyttjande av systemets cacheminne och därefter desto lägre fördröjning för deras bearbetning. Det finns dock många anledningar till att paketen inte kommer fram i en skur. En av huvudorsakerna som undersöks i detta projekt är paketinterfoliering som sker i routrarna och switcharna längs vägen som paketen tar från en dator till/från datacentret. Detta projekt tar en mer allmän titt på trafik som kommer till en server via ett träd av nätverksenheter på upplänksvägen till servern. Fokus är att använda skript för att skapa ett träd av switchar och utföra experiment med dem genom att skripta Mininet. Från dessa experiment lärde vi oss att de olika algoritmerna som styr utgångsköerna för nätverksnoderna över nätverksvägen spelar en betydande roll i paketinterfoliering. Dessutom förklarade experiment på Mininet-värdanslutning hur med Openflow-protokollet, Mininet-styrenheten satte upp regler i switcharna i nätverkstopologin. Slutligen visade experiment i TCP-genomströmning de begränsande faktorerna för en TCP-anslutning mellan servern och en värd medan många tillhandahållna trafikflöden illustrerar vanliga beteenden för paketinterfoliering som uppstår på grund av switcharna.

Children and young people rank among the most endangered social groups. Not only do they go through an important stage of personality formation which could be influenced by many factors, but this phase of personal development influences, to a certain extent, their future lives and the possibility of a fully-fledged social integration. At this stage, children and young people develop stronger perceptual sensitivity of either positive or negative outer factors. These factors can have a stronger impact on children and young people living in an environment which the majority society perceives not as standard, but as a socially excluded, and hence unwanted, locality. The group of children and young people, especially the ones from socially excluded localities, are very often affected by the use of addictive substances. There is a big risk, that members of the group will become regular users of addicted substances which generate further negative effects. The dissertation is focused on the use of addictive substances by children living in socially excluded localities. The aim of the thesis is not only to explain terms like a socially excluded locality, or addictive substance, but also to point out to an extent and kind of perception and knowledge of addictive substances on the part of the monitored group...

De nos jours, des processeurs à mémoire partagée cohérente ayant jusqu’à 100 coresintégrés sur la même puce sont une réalité et des processeurs many-cores ayant plusieurs centaines, voire, un millier de cores sont à prévoir prochainement.Dans ces architectures, la question de la localité du trafic lié aux miss decaches L1 (data, instruction et TLB) est primordiale à la fois pour passer àl’échelle et pour réduire la consommation électrique (énergie consommée par bittransféré). Notre thèse est que : (i) la gestion de la localité des accès mémoiredoit être prise en compte au niveau du noyau du système d’exploitation et elle doitêtre effectuée d’une manière transparente aux applications utilisateur; et (ii) les noyaux monolithiques actuels sont incapables de renforcer la localité des accès mémoire des threads d’une même application parallèle, car la notion de threadsdans ces noyaux est intrinsèquement inadaptée pour les processeurs many-cores.Par conséquent, nous pensons que la démarche suivie jusqu’à présent pour faireévoluer les noyaux monolithiques n’est pas suffisante et qu’il est impératif demettre la question de la localité des accès mémoire au centre de cette évolution.Pour prouver notre thèse, nous avons conçu et réalisé ALMOS (Advanced Locality Management Operating System), un système d’exploitation expérimental à base de noyau monolithique distribué. ALMOS dispose d’un nouveau concept de thread, nommé Processus Hybrides. Il permet à son noyau de renforcer, d’une manière transparente, la localité des accès mémoire liés à l'exécution de chaque thread. La gestion des ressources (cores et mémoires physiques) dans le noyau d’ALMOS est distribuée renforçant la localité des accès mémoire lors de la réalisation des services systèmes. La prise de décision concernant l’allocation mémoire, le placement des tâches et l’équilibrage de charge dans le noyau d’ALMOS est décentralisée, multi-critères et sans prise de verrou. Elle repose sur une infrastructure distribuée coordonnant d’une manière scalable l’accès aux ressources.En utilisant le prototype virtuel précis au cycle et au bit près du processeur many-core TSAR, nous avons expérimentalement démontré que : (i) les performances(scalabilité et temps d’exécution) du schéma d'ordonnancement distribué du noyaud’ALMOS sur 256 cores dépassent celles des noyaux monolithiques existants; (ii) la réalisation répartie de l’appel système fork permet de passer à l’échellece service système sur 512 cores; (iii) le coût de la mise à jour de l’infrastructure distribué de prise de décisions du noyau d’ALMOS ne nécessiteque 0.05% de la puissance de calcul totale du processeur TSAR; (iv) les performances(scalabilité, temps d’exécution et trafic distant) de la stratégie d’affinitémémoire du noyau d’ALMOS, nommé Auto-Next-Touch, dépassent celles des deuxstratégies First-Touch et Interleave sur 64 cores; (v) le modèle de processushybrides d’ALMOS permet de passer à l’échelle deux applications hautementmulti-threads existantes sur 256 cores et une troisième sur 1024 cores; et enfin (vi) le couple ALMOS/TSAR (64 cores) offre systématiquement une bien meilleure scalabilité que le couple Linux/AMD (Interlagos 64 cores) pour 8 applications de classe HPC et traitement d’images. / Nowadays, single-chip cache-coherent many-core processors having up to 100 coresare a reality. Many-cores with hundreds or even a thousand of cores are planned in the near future. In theses architectures, the question of the locality of L1 cache-miss related traffic (data, instruction and TLB) is essential for both scalability and power consumption (energy by moved bit). Our thesis is that: (i) handling the locality of memory accesses should be done at kernel level of an operating system in a transparent manner to user applications; and (ii) the current monolithic kernels are not able to enforce the locality of memory accesses of multi-threaded applications, because the concept of thread in these kernels is inherently unsuitable for many-core processors. Therefore, we believe that the evolution approach of monolithic kernels undertaken until now is insufficient and it is imperative to put the question of the locality of memory accesses in the heart of this evolution.To prove our thesis, we designed and implemented ALMOS (Advanced Locality Management Operating System), an experimental operating system based on a distributed monolithic kernel. ALMOS has a new concept of thread, called Hybrid Process. It allows its kernel to enforce the locality of memory accesses of each thread. The resources (cores and physical memory) management in ALMOS's kernel is distributed enforcing the locality of memory accesses when performing system services. Decision making regarding memory allocation, tasks placement and load balancing in ALMOS's kernel is decentralized, multi-criteria and without locking. It is based on a distributed infrastructure coordinating, in a scalable manner, the accesses to resources.Using the cycle accurate and bit accurate virtual prototype of TSAR many-core processor, we experimentally demonstrated that: (i) performance (scalability and execution time) on 256 cores of the distributed scheduling scheme of ALMOS's kernel outperform those of the shared scheduling scheme found in existing monolithic kernels; (ii) distributed realization of the fork system call enables this system service to scale on 512 cores; (iii) updating the distrusted decision-making infrastructure of ALMOS's kernel costs just 0.05 % of the total computing power of TSAR processor; (iv) performance (scalability, execution time and remote traffic) of memory affinity strategy of ALMOS's kernel, called Auto-Next-Touch, outperform those of two existing strategies First-Touch and Interleave on 64 cores; (v) concept of Hybrid Process of ALMOS's kernel scales up two existing highly multi-threads applications on 256 cores and a third one on 1024 cores; and finally (vi) the couple ALMOS/TSAR (64 cores) gives systematically much better scalability than the couple Linux/AMD (Interlagos 64 cores) for 8 multi-threads applications belonging to HPC and image processing domains.

Multi-armed bandits (MABs) have been studied extensively in the literature and have applications in a wealth of domains, including recommendation systems, dynamic pricing, and investment management. On the one hand, the current MAB literature largely seems to focus on the setting where each arm is available to play at each time step, and ignores how agents move between the arms. On the other hand, there is work that takes the movement between arms into account, but this work models the problem as a Markov decision process and applies generic reinforcement learning (RL) algorithms, like Q-learning. This thesis examines an extension of the MAB problem to a setting where the set of available arms at each round depends on which arm was played in the previous round. In this formulation the arms are nodes in a graph, and arms that can be played successively are connected via edges. We denote this the Graph Bandit (GB) problem. We show that under certain conditions the optimal action is governed by a stationary policy. Furthermore, we develop an algorithm that leverages the graphical structure of the problem to find this policy when the reward distributions are perfectly known, and denote this algorithm the Q-graph. When the reward distributions are unknown, we show how to leverage the Qgraph algorithm together with standard sampling algorithms like Thompson sampling and upper confidence bound to create an online learning algorithm that provably achieves logarithmic regret. Finally, this regret-bound is supported in numerical simulations, and it is illustrated how the proposed Q-graph algorithm outperforms generic algorithms from the MAB and RL communities. / Flerarmade banditer (FAB) har studerats omfattande i litteraturen och har applikationer inom en mängd domäner, såsom rekommendationssystem, dynamisk prissättning och finans. Å ena sidan verkar det som at en stor del av litteraturen fokuserar på situationen där alla armar är tillgängliga att spela vid varje tidssteg och ignorerar hur agenten rör sig mellan armarna. Å andra sidan finns det arbete som tar till hänsyn hur agenten rör sig mellan armarna men det arbetet modellerar systemet som en Markovprocess och använder sig av generiska inlärningsmetoder, såsom Q-learning. Den här uppsatsen undersöker en utvidgning av FAB-problemet till en situation där mängden tillgänliga armar vid varje runda beror på vilken arm som spelades i den föregående rundan. I denna formulering är armarna noder i en graf och armar som kan spelas i på varandra följande rundor är anslutna via kanter. Vi kallar det här problemt Grafbanditen. Vi visar att under vissa förutsättningar bestäms det optimala aggerandet av en stationär policy. Vi utvecklar också en algoritm som utnyttjar den grafiska strukturen i problemet för att beräkna denna policy när distributionerna hos alla armar är kända. Denna algoritm får namnet Q-grafen. När distributionerna är okända visar vi hur Q-grafen kan användas tillsammans med Thompson sampling eller upper confidence bound-metoder för att skapa en online inlärningsalgoritm som bevisligen uppnår logaritmisk regret. Slutligen stöds de teoretiska resultaten via numeriska simuleringar som illustrerar att Q-grafen är överlägsen många generiska inlärningsalgoritmer.

No description available.

No description available.

No description available.

No description available.

No description available.