On-line defragmentation for run-time partially reconfigurable FPGAs

Dynamically reconfigurable systems have benefited from a new class of FPGAs recently introduced into the market, which allow partial and dynamic reconfiguration at run-time, enabling multiple independent functions from different applications to share the same device, swapping resources as needed. When the sequence of tasks to be performed is not predictable, resource allocation decisions have to be made on-line, fragmenting the FPGA logic space. A rearrangement may be necessary to get enough contiguous space to efficiently implement incoming functions, to avoid spreading their components and, as a result, degrading their performance. This paper presents a novel active replication mechanism for configurable logic blocks (CLBs), able to implement on-line rearrangements, defragmenting the available FPGA resources without disturbing those functions that are currently running.

An Intelligent Distributed System for Environmental Management

Environmental management is a complex task. The amount and heterogeneity of the data needed for an environmental decision making tool is overwhelming without adequate database systems and innovative methodologies. As far as data management, data interaction and data processing is concerned we here propose the use of a Geographical Information System (GIS) whilst for the decision making we suggest a Multi-Agent System (MAS) architecture. With the adoption of a GIS we hope to provide a complementary coexistence between heterogeneous data sets, a correct data structure, a good storage capacity and a friendly user’s interface. By choosing a distributed architecture such as a Multi-Agent System, where each agent is a semi-autonomous Expert System with the necessary skills to cooperate with the others in order to solve a given task, we hope to ensure a dynamic problem decomposition and to achieve a better performance compared with standard monolithical architectures. Finally, and in view of the partial, imprecise, and ever changing character of information available for decision making, Belief Revision capabilities are added to the system. Our aim is to present and discuss an intelligent environmental management system capable of suggesting the more appropriate land-use actions based on the existing spatial and non-spatial constraints.

Comparison of single and multi-objective evolutionary algorithms for robust link-state routing

Traffic Engineering (TE) approaches are increasingly impor- tant in network management to allow an optimized configuration and resource allocation. In link-state routing, the task of setting appropriate weights to the links is both an important and a challenging optimization task. A number of different approaches has been put forward towards this aim, including the successful use of Evolutionary Algorithms (EAs). In this context, this work addresses the evaluation of three distinct EAs, a single and two multi-objective EAs, in two tasks related to weight setting optimization towards optimal intra-domain routing, knowing the network topology and aggregated traffic demands and seeking to mini- mize network congestion. In both tasks, the optimization considers sce- narios where there is a dynamic alteration in the state of the system, in the first considering changes in the traffic demand matrices and in the latter considering the possibility of link failures. The methods will, thus, need to simultaneously optimize for both conditions, the normal and the altered one, following a preventive TE approach towards robust configurations. Since this can be formulated as a bi-objective function, the use of multi-objective EAs, such as SPEA2 and NSGA-II, came nat- urally, being those compared to a single-objective EA. The results show a remarkable behavior of NSGA-II in all proposed tasks scaling well for harder instances, and thus presenting itself as the most promising option for TE in these scenarios.

Emergence of Cooperation on Static and Dynamic Networks

Game theory is a branch of applied mathematics used to analyze situation where two or more agents are interacting. Originally it was developed as a model for conflicts and collaborations between rational and intelligent individuals. Now it finds applications in social sciences, eco- nomics, biology (particularly evolutionary biology and ecology), engineering, political science, international relations, computer science, and philosophy. Networks are an abstract representation of interactions, dependencies or relationships. Net- works are extensively used in all the fields mentioned above and in many more. Many useful informations about a system can be discovered by analyzing the current state of a network representation of such system. In this work we will apply some of the methods of game theory to populations of agents that are interconnected. A population is in fact represented by a network of players where one can only interact with another if there is a connection between them. In the first part of this work we will show that the structure of the underlying network has a strong influence on the strategies that the players will decide to adopt to maximize their utility. We will then introduce a supplementary degree of freedom by allowing the structure of the population to be modified along the simulations. This modification allows the players to modify the structure of their environment to optimize the utility that they can obtain.

Hierarchical neural network with high storage capacity

A recent method used to optimize biased neural networks with low levels of activity is applied to a hierarchical model. As a consequence, the performance of the system is strongly enhanced. The steps to achieve optimization are analyzed in detail.

Video transcoding in a distributed cloud computing environment

Video transcoding refers to the process of converting a digital video from one format into another format. It is a compute-intensive operation. Therefore, transcoding of a large number of simultaneous video streams requires a large amount of computing resources. Moreover, to handle di erent load conditions in a cost-e cient manner, the video transcoding service should be dynamically scalable. Infrastructure as a Service Clouds currently offer computing resources, such as virtual machines, under the pay-per-use business model. Thus the IaaS Clouds can be leveraged to provide a coste cient, dynamically scalable video transcoding service. To use computing resources e ciently in a cloud computing environment, cost-e cient virtual machine provisioning is required to avoid overutilization and under-utilization of virtual machines. This thesis presents proactive virtual machine resource allocation and de-allocation algorithms for video transcoding in cloud computing. Since users' requests for videos may change at di erent times, a check is required to see if the current computing resources are adequate for the video requests. Therefore, the work on admission control is also provided. In addition to admission control, temporal resolution reduction is used to avoid jitters in a video. Furthermore, in a cloud computing environment such as Amazon EC2, the computing resources are more expensive as compared with the storage resources. Therefore, to avoid repetition of transcoding operations, a transcoded video needs to be stored for a certain time. To store all videos for the same amount of time is also not cost-e cient because popular transcoded videos have high access rate while unpopular transcoded videos are rarely accessed. This thesis provides a cost-e cient computation and storage trade-o strategy, which stores videos in the video repository as long as it is cost-e cient to store them. This thesis also proposes video segmentation strategies for bit rate reduction and spatial resolution reduction video transcoding. The evaluation of proposed strategies is performed using a message passing interface based video transcoder, which uses a coarse-grain parallel processing approach where video is segmented at group of pictures level.

Scheduling dynamic dataflow graphs with model checking

With the shift towards many-core computer architectures, dataflow programming has been proposed as one potential solution for producing software that scales to a varying number of processor cores. Programming for parallel architectures is considered difficult as the current popular programming languages are inherently sequential and introducing parallelism is typically up to the programmer. Dataflow, however, is inherently parallel, describing an application as a directed graph, where nodes represent calculations and edges represent a data dependency in form of a queue. These queues are the only allowed communication between the nodes, making the dependencies between the nodes explicit and thereby also the parallelism. Once a node have the su cient inputs available, the node can, independently of any other node, perform calculations, consume inputs, and produce outputs. Data ow models have existed for several decades and have become popular for describing signal processing applications as the graph representation is a very natural representation within this eld. Digital lters are typically described with boxes and arrows also in textbooks. Data ow is also becoming more interesting in other domains, and in principle, any application working on an information stream ts the dataflow paradigm. Such applications are, among others, network protocols, cryptography, and multimedia applications. As an example, the MPEG group standardized a dataflow language called RVC-CAL to be use within reconfigurable video coding. Describing a video coder as a data ow network instead of with conventional programming languages, makes the coder more readable as it describes how the video dataflows through the different coding tools. While dataflow provides an intuitive representation for many applications, it also introduces some new problems that need to be solved in order for data ow to be more widely used. The explicit parallelism of a dataflow program is descriptive and enables an improved utilization of available processing units, however, the independent nodes also implies that some kind of scheduling is required. The need for efficient scheduling becomes even more evident when the number of nodes is larger than the number of processing units and several nodes are running concurrently on one processor core. There exist several data ow models of computation, with different trade-offs between expressiveness and analyzability. These vary from rather restricted but statically schedulable, with minimal scheduling overhead, to dynamic where each ring requires a ring rule to evaluated. The model used in this work, namely RVC-CAL, is a very expressive language, and in the general case it requires dynamic scheduling, however, the strong encapsulation of dataflow nodes enables analysis and the scheduling overhead can be reduced by using quasi-static, or piecewise static, scheduling techniques. The scheduling problem is concerned with nding the few scheduling decisions that must be run-time, while most decisions are pre-calculated. The result is then an, as small as possible, set of static schedules that are dynamically scheduled. To identify these dynamic decisions and to find the concrete schedules, this thesis shows how quasi-static scheduling can be represented as a model checking problem. This involves identifying the relevant information to generate a minimal but complete model to be used for model checking. The model must describe everything that may affect scheduling of the application while omitting everything else in order to avoid state space explosion. This kind of simplification is necessary to make the state space analysis feasible. For the model checker to nd the actual schedules, a set of scheduling strategies are de ned which are able to produce quasi-static schedulers for a wide range of applications. The results of this work show that actor composition with quasi-static scheduling can be used to transform data ow programs to t many different computer architecture with different type and number of cores. This in turn, enables dataflow to provide a more platform independent representation as one application can be fitted to a specific processor architecture without changing the actual program representation. Instead, the program representation is in the context of design space exploration optimized by the development tools to fit the target platform. This work focuses on representing the dataflow scheduling problem as a model checking problem and is implemented as part of a compiler infrastructure. The thesis also presents experimental results as evidence of the usefulness of the approach.

Cost-efficient virtual machine management : provisioning, admission control, and consolidation

One of the main challenges in Software Engineering is to cope with the transition from an industry based on software as a product to software as a service. The field of Software Engineering should provide the necessary methods and tools to develop and deploy new cost-efficient and scalable digital services. In this thesis, we focus on deployment platforms to ensure cost-efficient scalability of multi-tier web applications and on-demand video transcoding service for different types of load conditions. Infrastructure as a Service (IaaS) clouds provide Virtual Machines (VMs) under the pay-per-use business model. Dynamically provisioning VMs on demand allows service providers to cope with fluctuations on the number of service users. However, VM provisioning must be done carefully, because over-provisioning results in an increased operational cost, while underprovisioning leads to a subpar service. Therefore, our main focus in this thesis is on cost-efficient VM provisioning for multi-tier web applications and on-demand video transcoding. Moreover, to prevent provisioned VMs from becoming overloaded, we augment VM provisioning with an admission control mechanism. Similarly, to ensure efficient use of provisioned VMs, web applications on the under-utilized VMs are consolidated periodically. Thus, the main problem that we address is cost-efficient VM provisioning augmented with server consolidation and admission control on the provisioned VMs. We seek solutions for two types of applications: multi-tier web applications that follow the request-response paradigm and on-demand video transcoding that is based on video streams with soft realtime constraints. Our first contribution is a cost-efficient VM provisioning approach for multi-tier web applications. The proposed approach comprises two subapproaches: a reactive VM provisioning approach called ARVUE and a hybrid reactive-proactive VM provisioning approach called Cost-efficient Resource Allocation for Multiple web applications with Proactive scaling. Our second contribution is a prediction-based VM provisioning approach for on-demand video transcoding in the cloud. Moreover, to prevent virtualized servers from becoming overloaded, the proposed VM provisioning approaches are augmented with admission control approaches. Therefore, our third contribution is a session-based admission control approach for multi-tier web applications called adaptive Admission Control for Virtualized Application Servers. Similarly, the fourth contribution in this thesis is a stream-based admission control and scheduling approach for on-demand video transcoding called Stream-Based Admission Control and Scheduling. Our fifth contribution is a computation and storage trade-o strategy for cost-efficient video transcoding in cloud computing. Finally, the sixth and the last contribution is a web application consolidation approach, which uses Ant Colony System to minimize the under-utilization of the virtualized application servers.

Using genetic algorithms for the single allocation hub location problem

Hub location problem is an NP-hard problem that frequently arises in the design of transportation and distribution systems, postal delivery networks, and airline passenger flow. This work focuses on the Single Allocation Hub Location Problem (SAHLP). Genetic Algorithms (GAs) for the capacitated and uncapacitated variants of the SAHLP based on new chromosome representations and crossover operators are explored. The GAs is tested on two well-known sets of real-world problems with up to 200 nodes. The obtained results are very promising. For most of the test problems the GA obtains improved or best-known solutions and the computational time remains low. The proposed GAs can easily be extended to other variants of location problems arising in network design planning in transportation systems.

An Interactive Theorem Prover for First-Order Dynamic Logic

Dynamic logic is an extension of modal logic originally intended for reasoning about computer programs. The method of proving correctness of properties of a computer program using the well-known Hoare Logic can be implemented by utilizing the robustness of dynamic logic. For a very broad range of languages and applications in program veri cation, a theorem prover named KIV (Karlsruhe Interactive Veri er) Theorem Prover has already been developed. But a high degree of automation and its complexity make it di cult to use it for educational purposes. My research work is motivated towards the design and implementation of a similar interactive theorem prover with educational use as its main design criteria. As the key purpose of this system is to serve as an educational tool, it is a self-explanatory system that explains every step of creating a derivation, i.e., proving a theorem. This deductive system is implemented in the platform-independent programming language Java. In addition, a very popular combination of a lexical analyzer generator, JFlex, and the parser generator BYacc/J for parsing formulas and programs has been used.

Multi-Objective Genetic Algorithms for the Single Allocation Hub Location Problem

Hub Location Problems play vital economic roles in transportation and telecommunication networks where goods or people must be efficiently transferred from an origin to a destination point whilst direct origin-destination links are impractical. This work investigates the single allocation hub location problem, and proposes a genetic algorithm (GA) approach for it. The effectiveness of using a single-objective criterion measure for the problem is first explored. Next, a multi-objective GA employing various fitness evaluation strategies such as Pareto ranking, sum of ranks, and weighted sum strategies is presented. The effectiveness of the multi-objective GA is shown by comparison with an Integer Programming strategy, the only other multi-objective approach found in the literature for this problem. Lastly, two new crossover operators are proposed and an empirical study is done using small to large problem instances of the Civil Aeronautics Board (CAB) and Australian Post (AP) data sets.

A Study of Ordered Gene Problems Featuring DNA Error Correction and DNA Fragment Assembly with a Variety of Heuristics, Genetic Algorithm Variations, and Dynamic Representations

Ordered gene problems are a very common classification of optimization problems. Because of their popularity countless algorithms have been developed in an attempt to find high quality solutions to the problems. It is also common to see many different types of problems reduced to ordered gene style problems as there are many popular heuristics and metaheuristics for them due to their popularity. Multiple ordered gene problems are studied, namely, the travelling salesman problem, bin packing problem, and graph colouring problem. In addition, two bioinformatics problems not traditionally seen as ordered gene problems are studied: DNA error correction and DNA fragment assembly. These problems are studied with multiple variations and combinations of heuristics and metaheuristics with two distinct types or representations. The majority of the algorithms are built around the Recentering- Restarting Genetic Algorithm. The algorithm variations were successful on all problems studied, and particularly for the two bioinformatics problems. For DNA Error Correction multiple cases were found with 100% of the codes being corrected. The algorithm variations were also able to beat all other state-of-the-art DNA Fragment Assemblers on 13 out of 16 benchmark problem instances.

Characterizing Dynamic Optimization Benchmarks for the Comparison of Multi-Modal Tracking Algorithms

Population-based metaheuristics, such as particle swarm optimization (PSO), have been employed to solve many real-world optimization problems. Although it is of- ten sufficient to find a single solution to these problems, there does exist those cases where identifying multiple, diverse solutions can be beneficial or even required. Some of these problems are further complicated by a change in their objective function over time. This type of optimization is referred to as dynamic, multi-modal optimization. Algorithms which exploit multiple optima in a search space are identified as niching algorithms. Although numerous dynamic, niching algorithms have been developed, their performance is often measured solely on their ability to find a single, global optimum. Furthermore, the comparisons often use synthetic benchmarks whose landscape characteristics are generally limited and unknown. This thesis provides a landscape analysis of the dynamic benchmark functions commonly developed for multi-modal optimization. The benchmark analysis results reveal that the mechanisms responsible for dynamism in the current dynamic bench- marks do not significantly affect landscape features, thus suggesting a lack of representation for problems whose landscape features vary over time. This analysis is used in a comparison of current niching algorithms to identify the effects that specific landscape features have on niching performance. Two performance metrics are proposed to measure both the scalability and accuracy of the niching algorithms. The algorithm comparison results demonstrate the algorithms best suited for a variety of dynamic environments. This comparison also examines each of the algorithms in terms of their niching behaviours and analyzing the range and trade-off between scalability and accuracy when tuning the algorithms respective parameters. These results contribute to the understanding of current niching techniques as well as the problem features that ultimately dictate their success.

Planification et dimensionnement des réseaux optiques de longues distances

Le projet de recherche porte sur l'étude des problèmes de conception et de planification d'un réseau optique de longue distance, aussi appelé réseau de coeur (OWAN-Optical Wide Area Network en anglais). Il s'agit d'un réseau qui transporte des flots agrégés en mode commutation de circuits. Un réseau OWAN relie différents sites à l'aide de fibres optiques connectées par des commutateurs/routeurs optiques et/ou électriques. Un réseau OWAN est maillé à l'échelle d'un pays ou d’un continent et permet le transit des données à très haut débit. Dans une première partie du projet de thèse, nous nous intéressons au problème de conception de réseaux optiques agiles. Le problème d'agilité est motivé par la croissance de la demande en bande passante et par la nature dynamique du trafic. Les équipements déployés par les opérateurs de réseaux doivent disposer d'outils de configuration plus performants et plus flexibles pour gérer au mieux la complexité des connexions entre les clients et tenir compte de la nature évolutive du trafic. Souvent, le problème de conception d'un réseau consiste à prévoir la bande passante nécessaire pour écouler un trafic donné. Ici, nous cherchons en plus à choisir la meilleure configuration nodale ayant un niveau d'agilité capable de garantir une affectation optimale des ressources du réseau. Nous étudierons également deux autres types de problèmes auxquels un opérateur de réseau est confronté. Le premier problème est l'affectation de ressources du réseau. Une fois que l'architecture du réseau en termes d'équipements est choisie, la question qui reste est de savoir : comment dimensionner et optimiser cette architecture pour qu'elle rencontre le meilleur niveau possible d'agilité pour satisfaire toute la demande. La définition de la topologie de routage est un problème d'optimisation complexe. Elle consiste à définir un ensemble de chemins optiques logiques, choisir les routes physiques suivies par ces derniers, ainsi que les longueurs d'onde qu'ils utilisent, de manière à optimiser la qualité de la solution obtenue par rapport à un ensemble de métriques pour mesurer la performance du réseau. De plus, nous devons définir la meilleure stratégie de dimensionnement du réseau de façon à ce qu'elle soit adaptée à la nature dynamique du trafic. Le second problème est celui d'optimiser les coûts d'investissement en capital(CAPEX) et d'opération (OPEX) de l'architecture de transport proposée. Dans le cas du type d'architecture de dimensionnement considérée dans cette thèse, le CAPEX inclut les coûts de routage, d'installation et de mise en service de tous les équipements de type réseau installés aux extrémités des connexions et dans les noeuds intermédiaires. Les coûts d'opération OPEX correspondent à tous les frais liés à l'exploitation du réseau de transport. Étant donné la nature symétrique et le nombre exponentiel de variables dans la plupart des formulations mathématiques développées pour ces types de problèmes, nous avons particulièrement exploré des approches de résolution de type génération de colonnes et algorithme glouton qui s'adaptent bien à la résolution des grands problèmes d'optimisation. Une étude comparative de plusieurs stratégies d'allocation de ressources et d'algorithmes de résolution, sur différents jeux de données et de réseaux de transport de type OWAN démontre que le meilleur coût réseau est obtenu dans deux cas : une stratégie de dimensionnement anticipative combinée avec une méthode de résolution de type génération de colonnes dans les cas où nous autorisons/interdisons le dérangement des connexions déjà établies. Aussi, une bonne répartition de l'utilisation des ressources du réseau est observée avec les scénarios utilisant une stratégie de dimensionnement myope combinée à une approche d'allocation de ressources avec une résolution utilisant les techniques de génération de colonnes. Les résultats obtenus à l'issue de ces travaux ont également démontré que des gains considérables sont possibles pour les coûts d'investissement en capital et d'opération. En effet, une répartition intelligente et hétérogène de ressources d’un réseau sur l'ensemble des noeuds permet de réaliser une réduction substantielle des coûts du réseau par rapport à une solution d'allocation de ressources classique qui adopte une architecture homogène utilisant la même configuration nodale dans tous les noeuds. En effet, nous avons démontré qu'il est possible de réduire le nombre de commutateurs photoniques tout en satisfaisant la demande de trafic et en gardant le coût global d'allocation de ressources de réseau inchangé par rapport à l'architecture classique. Cela implique une réduction substantielle des coûts CAPEX et OPEX. Dans nos expériences de calcul, les résultats démontrent que la réduction de coûts peut atteindre jusqu'à 65% dans certaines jeux de données et de réseau.

A layered JavaScript virtual machine supporting dynamic instrumentation

L’observation de l’exécution d’applications JavaScript est habituellement réalisée en instrumentant une machine virtuelle (MV) industrielle ou en effectuant une traduction source-à-source ad hoc et complexe. Ce mémoire présente une alternative basée sur la superposition de machines virtuelles. Notre approche consiste à faire une traduction source-à-source d’un programme pendant son exécution pour exposer ses opérations de bas niveau au travers d’un modèle objet flexible. Ces opérations de bas niveau peuvent ensuite être redéfinies pendant l’exécution pour pouvoir en faire l’observation. Pour limiter la pénalité en performance introduite, notre approche exploite les opérations rapides originales de la MV sous-jacente, lorsque cela est possible, et applique les techniques de compilation à-la-volée dans la MV superposée. Notre implémentation, Photon, est en moyenne 19% plus rapide qu’un interprète moderne, et entre 19× et 56× plus lente en moyenne que les compilateurs à-la-volée utilisés dans les navigateurs web populaires. Ce mémoire montre donc que la superposition de machines virtuelles est une technique alternative compétitive à la modification d’un interprète moderne pour JavaScript lorsqu’appliqué à l’observation à l’exécution des opérations sur les objets et des appels de fonction.