No description available.

No description available.

Quality-of-Service (QoS) to users is a critical requirement of resource allocation in wireless networks and has drawn significant research attention over a long time. However, the QoS requirements differ vastly based on the wireless network paradigm. At one extreme, we have a millimeter wave small-cell network for streaming data that requires very high throughput and low latency. At the other end, we have Machine-to-Machine (M2M) uplink traffic with low throughput and low latency. In this dissertation, we investigate and solve QoS-aware resource allocation problems for diverse wireless paradigms. We first study cross-layer dynamic spectrum allocation in a LTE macro-cellular network with fractional frequency reuse to improve the spectral efficiency for cell-edge users. We show that the resultant optimization problem is NP-hard and propose a low-complexity layered spectrum allocation heuristic that strikes a balance between rate maximization and fairness of allocation. Next, we develop an energy efficient downlink power control scheme in a energy harvesting small-cell base station equipped with local cache and wireless backhaul. We also study the tradeoff between the cache size and the energy harvesting capabilities. We next analyzed the file read latency in Distributed Storage Systems (DSS). We propose a heterogeneous DSS model wherein the stored data is categorized into multiple classes based on arrival rate of read requests, fault-tolerance for storage etc. Using a queuing theoretic approach, we establish bounds on the average read latency for different scheduling policies. We also show that erasure coding in DSS serves the dual purpose of reducing read latency and increasing the energy efficiency. Lastly, we investigate the problem of delay-efficient packet scheduling in M2M uplink with heterogeneous traffic characteristics. We classify the uplink traffic into multiple classes and propose a proportionally-fair delay-efficient heuristic packet scheduler. Using a queuing theoretic approach, we next develop a delay optimal multiclass packet scheduler and later extend it to joint medium access control and packet scheduling for M2M uplink. Using extensive simulations, we show that the proposed schedulers perform better than state-of-the-art schedulers in terms of average delay and packet delay jitter. / PHD

Electrical power systems consist of a large number of power generators connected to consumers through a complex system of transmission and distribution lines. Within the electric grid, a continuous balance between generation and consumption of electricity must be maintained., ensuring stable operation of the grid. In recent decades due to increasing electricity demand, there is an increased likelihood of electrical power systems experiencing stress conditions. These conditions lead to a limited supply and cascading failures throughout the grid that could lead to wide area outages. Demand Response (DR) is a method involving the curtailment of loads during critical peak load hours, that restores that balance between demand and supply of electricity. In order to implement DR and ensure efficient energy operation of buildings, detailed energy monitoring is essential. This information can then be used for energy management, by monitoring the power consumption of devices and giving users detailed feedback at an individual device level. Based on the data from the Energy Information Administration (EIA), approximately half of all commercial buildings in the U.S. are 5,000 square feet or smaller in size, whereas the majority of the rest is made up of medium-sized commercial buildings ranging in size between 5,001 and 50,000 square feet. Given that these medium-size buildings account for a large portion of the total energy demand, these buildings are an ideal target for participating in DR. In this dissertation, two broad solutions for commercial building DR have been presented. The first is a load disaggregation technique to disaggregate the power of individual HVACs using machine learning classification techniques, where a single power meter is used to collect aggregated HVAC power data of a building. This method is then tested over a number of case studies, from which it is found that the aggregated power data can be disaggregated to accurately predict the power consumption and state of activity of individual HVAC loads. The second work focuses on a DR algorithm involving the determination of an optimal bid price for double auctioning between the user and the electric utility, in addition to a load scheduling algorithm that controls single floor HVAC and lighting loads in a commercial building, considering user preferences and load priorities. A number of case studies are carried out, from which it is observed that the algorithm can effectively control loads within a given demand limit, while efficiently maintaining user preferences for a number of different load configurations and scenarios. Therefore, the major contributions of this work include- A novel HVAC power disaggregation technique using machine learning methods, and also a DR algorithm for HVAC and lighting load control, incorporating user preferences and load priorities based on a double-auction approach. / Doctor of Philosophy / Electrical power systems consist of a large number of power generators connected to consumers through a complex system of transmission and distribution lines. Within the electric grid, a continuous balance between generation and consumption of electricity must be maintained., ensuring stable operation of the grid. When electricity demand is high, Demand Response (DR) is a method that can be used to reduce user loads, restoring the balance between demand and supply of electricity. Based on data from the Energy Information Administration (EIA), half of all commercial buildings in the US measure 5,000 square feet or smaller in size, whereas the majority of the other half is made up of medium-sized commercial buildings measuring in at between 5,001 to 50,000 square feet. This makes these commercial buildings an ideal target for participating in DR. In this dissertation, two broad solutions for commercial building DR have been presented. The first is a load disaggregation technique, where power consumption and activity of individual HVACs can be obtained, using a single power meter. The second work focuses on a DR algorithm, that controls single floor HVAC and lighting loads in a commercial building, based on a user generated bid price for electricity, user preferences and load priorities, when electricity demand is at its peak.

Virtualization is one of the important enabling technologies for Cloud Computing which facilitates sharing of resources among the virtual machines. However, it incurs performance overheads due to contention of physical devices such as disk and network bandwidth. Various I/O applications having different latency requirements may be executing concurrently on different virtual machines provisioned on a single server in Cloud data-centers. It is pertinent that the performance SLAs of such applications are satisfied through intelligent scheduling and allocation of disk resources. The underlying disk scheduler at the server is unable to distinguish between the application requests being oblivious to the characteristics of these applications. Therefore, all the applica- tions are provided best effort services by default. This may lead to performance degradation for the latency sensitive applications. In this work, we propose a novel disk scheduling framework PriDyn (Dynamic Priority) which provides differentiated services to various I/O applications co-located on a single host based on their latency attributes and desired performance. The framework employs a scheduling algorithm which dynamically computes latency estimates for all concurrent I/O applications for a given system state. Based on these, an appropriate pri- ority assignment for the applications is determined which is taken into consideration by the underlying disk scheduler at the host while scheduling the I/O applications on the physical disk. The proposed scheduling framework is able to successfully satisfy QoS requirements for the concurrent I/O applications within system constraints. This has been verified through ex- tensive experimental analysis. In order to realize the benefits of differentiated services provided by the PriDyn scheduler, proper combination of I/O applications must be ensured for the servers through intelligent meta-scheduling techniques at the Cloud data-center level. For achieving this, in the second part of this work, we extended the PriDyn framework to design a proactive admission control and scheduling framework PCOS (P rescient C loud I/O S cheduler). It aims to maximize to Utilization of disk resources without adversely affecting the performance of the applications scheduled on the systems. By anticipating the performance of the systems running multiple I/O applications, PCOS prevents the scheduling of undesirable workloads on them in order to maintain the necessary balance between resource consolidation and application performance guarantees. The PCOS framework includes the PriDyn scheduler as an important component and utilizes the dynamic disk resource allocation capabilities of PriDyn for meeting its goals. Experimental validations performed on real world I/O traces demonstrate that the proposed framework achieves appreciable enhancements in I/O performance through selection of optimal I/O workload combinations, indicating that this approach is a promising step towards enabling QoS guarantees for Cloud data-centers.

Approved for public release; distribution is unlimited / The Computer Aided Prototyping System (CAPS) and the Prototype System Description Language (PSDL) represent a pioneering effort in the field of software development. The implementation of CAPS will enable software engineers to automatically validate design specifications and functional requirements early in the design of a software system through the development and execution of a prototype of the system under construction. Execution of the prototype is controlled by an Execution Support System (ESS) within the framework of CAPS. One of the critical elements of the ESS is the Static Scheduler which extracts critical timing constraints and precedence information about operators from the PSDL source that describes the prototype. The Static Scheduler then uses this information to determine whether a feasible schedule can be built, and if it can, constructs the schedule for operator execution within the prototype. / http://archive.org/details/staticschedulerf00marl / Lieutenant Commander, United States Navy

O planejamento de lavra de curto prazo é normalmente executado utilizando-se número reduzido de informações de sondagem. Para aprimorar o gerenciamento de riscos geológicos no planejamento de lavra de curto prazo é necessário utilizar um universo maior de informações. Como é normalmente inviável obter novas informações de sondagem, esta dissertação propõe uma metodologia de utilização de amostras de pó de perfuratriz (a partir de furos de desmonte) como uma fonte de informação secundária e assim aprimorar a qualidade das estimativas. Neste sentido, foi adotada uma técnica de co-estimativa da variável P2O5 das sondagens (variável primária) em conjunto com a variável P2O5 do desmonte (variável secundária) baseado no modelo Marcoviano MM2, pelo qual é possível combinar as duas informações (sondagem e desmonte) na estimativa de um modelo de blocos. Este processo permitiu a modelagem de atributos geológicos de forma mais detalhada o que contribuiu para uma melhor interface entre o planejamento de curto prazo e a operação da mina. A metodologia proposta também possibilitou acessar uma população maior de informações geológicas o que contribui para a criação de planos operacionais mais aderentes aos objetivos de produção mensal ou semanal, e, ao mesmo tempo, respeitando o sequenciamento importado do planejamento de longo e médio prazo. Como resultado, foi demonstrado que é possível criar programas operacionais mais precisos com base em estimativas de áreas próximas à lavra mesmo quando apenas um pequeno número de informações primárias (sondagens) esteja disponível. / Short-term mine planning is normally carried out over a limited number of drillhole information. In order to improve the management of geological risks in mine planning, a larger population of samples is required. However, it is normally very difficult to obtain additional drillhole samples once mining takes place. This research addresses that issue by proposing a methodology for the incorporation of additional information from blastholes (secondary information) to the original drillhole samples (primary information). A co-estimation technique for using P2O5 samples from the drillholes (primary variable) in conjunction with P2O5 from blastholes (secondary variable) based on the Markovian estimation model (MM2), through which is possible to combine both sources of information for a better estimation of mineable blocks. This process has allowed more detailed modeling of geological attributes and a better interface between short-term mine planning and mine operations. The proposed methodology also allowed the access to a larger sample population which meant more accurate mine plans for the daily and weekly mine schedules. As a result, it has been demonstrated that it is possible to crate operational plans that are more precise through the use of models that are properly estimated even in those areas where only a small amount of primary information (drillhole samples) is available.

En informatique, la notion de processus léger ou thread est désormais omniprésente. En effet, les threads permettent à un programme d'une part d'exploiter pleinement les ordinateurs multiprocesseurs et d'autre part d'exprimer son parallélisme intrinsèque. Dans le domaine du calcul hautes performances, les threads permettent de recouvrir des communications ou plus généralement des entrées/sorties avec du calcul. Ils permettent aussi aux divers composants de l'application de progresser indépendamment les uns des autres, ce qui est nécessaire avec l'utilisation d'environnements de programmation toujours plus complexes comme MPI ou Corba.<br /><br />Mes travaux avaient pour objectif principal d'aboutir à la conception d'une bibliothèque de threads performante sur une vaste gamme d'architectures (machines mono ou multiprocesseurs, technologie "multithreading simultané", ...) et capable d'offrir les fonctionnalités demandées par les programmes de calcul hautes performances. Dans un premier temps, j'ai proposé une extension du modèle des Scheduler Activations pour le noyau Linux permettant d'être réactif aux interruptions dans une bibliothèque de threads utilisateurs. J'ai ensuite étendu ce mécanisme de manière à unifier la gestion des interruptions et des scrutations dans un environnement multithreadé. Enfin, j'ai proposé un mécanisme de prise de traces permettant de reconstituer précisément le déroulement d'un programme multithreadé, y compris lorsque l'ordonnancement est à deux niveaux.<br /><br />Ces travaux ont été implémentés au sein du logiciel PM2. La bibliothèque Marcel offre des threads performants sur une vaste gamme de processeurs et de systèmes en restant suffisamment flexible pour permettre aux applications qui le nécessitent de diriger précisément l'ordonnancement de leurs threads. Les applications peuvent être tracées de manière à pouvoir reconstituer et observer leur comportement précis. La trace obtenue peut être convertie au format du logiciel Pajé qui permet alors de visualiser graphiquement le déroulement de l'application.

This thesis explores memory performance for smartphone workloads. We design a Video Conference Workload (VCW) to model typical smartphone usage. We describe a trace-based methodology which uses a software implementation to mimic the behaviour of specialised hardware accelerators. Our methodology stores dataflow information from the original application to maintain the relationships between requests. We first study seven address mapping schemes with our VCW, using a first-ready, first-come-first-served (FR-FCFS) memory scheduler. Our results show the best performing scheme is up to 82% faster than the worst. The VCW is memory intensive, with up to 86.8% bandwidth utilisation using the best performing scheme. We also test a Web Browsing and a set of computer vision workloads. Most are not memory intensive, with utilisation under 15%. Finally, we compare four schedulers and find that the FR-FCFS scheduler using the Write Drain mode [8] performed the best, outperforming the worst scheduler by 6.3%.

Οι κατανεμημένοι αλγόριθμοι δρομολόγησης για switches βασισμένα σε κελιά (=πακέτα σταθερού μεγέθους - π.χ. ΑΤΜ) που χρησιμοποιούν την αρχιτεκτονική Virtual Output Queueing έχουν οδηγήσει στη κατασκευή δρομολογητών υψηλών επιδόσεων. Στην παρούσα διπλωματική παρουσιάζουμε μια νέα ιδέα για την κατασκευή τέτοιων αλγορίθμων που λέγεται "αμοιβαία προτεραιότητα". Οι αλγόριθμοι της οικογένειας αυτής, που λέγεται οικογένεια "MPE", πετυχαίνουν βέλτιστη εγγύηση εξυπηρέτησης για κάθε ουρά εισόδου (καθυστέρηση το πολύ N κύκλων για switch ΝχΝ). Επίσης όσον αφορά τη μέση καθυστέρηση των πακέτων , ξεπερνούν σε απόδοση όλους τους γνωστούς αλγορίθμους. Οι βελτιώσεις που επιτυγχάνονται είναι αρκετά σημαντικές (π.χ. μέχρι και 78% μικρότερη καθυστέρηση πακέτων σε σύγκριση με τον καλύτερο υπάρχοντα αλγόριθμο, για μια επανάληψη). / Distributed scheduling algorithms for cell-based (e.g. ATM) Virtual Output Queuing (VOQ) switches have led to high performance schedulers. In this thesis we introduce a new concept in making such algorithms called "mutual priority" . Algorithms of this family , called "MPE" , achieve optimal service guarantee (N in a NxN switch) and lead to packet delays that outperform all existing algorithms. The improvements achieved are quite significant (i.e. up to 78% less packet delay compared to the best existing algorithm for a single iteration).