Memory-based scheduling of scientific computing clusters

This study looks at how increased memory utilisation affects throughput and energy consumption in scientific computing, especially in high-energy physics. Our aim is to minimise energy consumed by a set of jobs without increasing the processing time. The earlier tests indicated that, especially in data analysis, throughput can increase over 100% and energy consumption decrease 50% by processing multiple jobs in parallel per CPU core. Since jobs are heterogeneous, it is not possible to find an optimum value for the number of parallel jobs. A better solution is based on memory utilisation, but finding an optimum memory threshold is not straightforward. Therefore, a fuzzy logic-based algorithm was developed that can dynamically adapt the memory threshold based on the overall load. In this way, it is possible to keep memory consumption stable with different workloads while achieving significantly higher throughput and energy-efficiency than using a traditional fixed number of jobs or fixed memory threshold approaches.

Empirics of the international inequality in CO2 emissions intensity: explanatory factors according to complementary decomposition methodologies

This paper analyses the international inequalities in CO2 emissions intensity for the period 1971–2009 and assesses explanatory factors. Multiplicative, group and additive methodologies of inequality decomposition are employed. The first allows us to clarify the separated role of the carbonisation index and the energy intensity in the pattern observed for inequalities in CO2 intensities; the second allows us to understand the role of regional groups; and the third allows us to investigate the role of different fossil energy sources (coal, oil and gas). The results show that, first, the reduction in global emissions intensity has coincided with a significant reduction in international inequality. Second, the bulk of this inequality and its reduction are attributed to differences between the groups of countries considered. Third, coal is the main energy source explaining these inequalities, although the growth in the relative contribution of gas is also remarkable. Fourth, the bulk of inequalities between countries and its decline are explained by differences in energy intensities, although there are significant differences in the patterns demonstrated by different groups of countries. JEL codes: D39; Q43; Q56. Key words: CO2 international distribution, inequality decomposition, CO2 emissions intensity

The M-Coffee web server: a meta-method for computing multiple sequence alignments by combining alternative alignment methods.

The M-Coffee server is a web server that makes it possible to compute multiple sequence alignments (MSAs) by running several MSA methods and combining their output into one single model. This allows the user to simultaneously run all his methods of choice without having to arbitrarily choose one of them. The MSA is delivered along with a local estimation of its consistency with the individual MSAs it was derived from. The computation of the consensus multiple alignment is carried out using a special mode of the T-Coffee package [Notredame, Higgins and Heringa (T-Coffee: a novel method for fast and accurate multiple sequence alignment. J. Mol. Biol. 2000; 302: 205-217); Wallace, O'Sullivan, Higgins and Notredame (M-Coffee: combining multiple sequence alignment methods with T-Coffee. Nucleic Acids Res. 2006; 34: 1692-1699)] Given a set of sequences (DNA or proteins) in FASTA format, M-Coffee delivers a multiple alignment in the most common formats. M-Coffee is a freeware open source package distributed under a GPL license and it is available either as a standalone package or as a web service from www.tcoffee.org.

Trust region versus line search for computing the optical flow

We consider the numerical treatment of the optical flow problem by evaluating the performance of the trust region method versus the line search method. To the best of our knowledge, the trust region method is studied here for the first time for variational optical flow computation. Four different optical flow models are used to test the performance of the proposed algorithm combining linear and nonlinear data terms with quadratic and TV regularization. We show that trust region often performs better than line search; especially in the presence of non-linearity and non-convexity in the model.

Effects of computing parameters and measurement locations on the estimation of 3D NPS in non-stationary MDCT images.

The goal of this study was to investigate the impact of computing parameters and the location of volumes of interest (VOI) on the calculation of 3D noise power spectrum (NPS) in order to determine an optimal set of computing parameters and propose a robust method for evaluating the noise properties of imaging systems. Noise stationarity in noise volumes acquired with a water phantom on a 128-MDCT and a 320-MDCT scanner were analyzed in the spatial domain in order to define locally stationary VOIs. The influence of the computing parameters in the 3D NPS measurement: the sampling distances bx,y,z and the VOI lengths Lx,y,z, the number of VOIs NVOI and the structured noise were investigated to minimize measurement errors. The effect of the VOI locations on the NPS was also investigated. Results showed that the noise (standard deviation) varies more in the r-direction (phantom radius) than z-direction plane. A 25 × 25 × 40 mm(3) VOI associated with DFOV = 200 mm (Lx,y,z = 64, bx,y = 0.391 mm with 512 × 512 matrix) and a first-order detrending method to reduce structured noise led to an accurate NPS estimation. NPS estimated from off centered small VOIs had a directional dependency contrary to NPS obtained from large VOIs located in the center of the volume or from small VOIs located on a concentric circle. This showed that the VOI size and location play a major role in the determination of NPS when images are not stationary. This study emphasizes the need for consistent measurement methods to assess and compare image quality in CT.

Evaluation of performance analysis software for Linux computing cluster

Tietokonejärjestelmän osien ja ohjelmistojen suorituskykymittauksista saadaan tietoa,jota voidaan käyttää suorituskyvyn parantamiseen ja laitteistohankintojen päätöksen tukena. Tässä työssä tutustutaan suorituskyvyn mittaamiseen ja mittausohjelmiin eli ns. benchmark-ohjelmistoihin. Työssä etsittiin ja arvioitiin eri tyyppisiä vapaasti saatavilla olevia benchmark-ohjelmia, jotka soveltuvat Linux-laskentaklusterin suorituskyvynanalysointiin. Benchmarkit ryhmiteltiin ja arvioitiin testaamalla niiden ominaisuuksia Linux-klusterissa. Työssä käsitellään myös mittausten tekemisen ja rinnakkaislaskennan haasteita. Benchmarkkeja löytyi moneen tarkoitukseen ja ne osoittautuivat laadultaan ja laajuudeltaan vaihteleviksi. Niitä on myös koottu ohjelmistopaketeiksi, jotta laitteiston suorituskyvystä saisi laajemman kuvan kuin mitä yhdellä ohjelmalla on mahdollista saada. Olennaista on ymmärtää nopeus, jolla dataa saadaan siirretyä prosessorille keskusmuistista, levyjärjestelmistä ja toisista laskentasolmuista. Tyypillinen benchmark-ohjelma sisältää paljon laskentaa tarvitsevan matemaattisen algoritmin, jota käytetään tieteellisissä ohjelmistoissa. Benchmarkista riippuen tulosten ymmärtäminen ja hyödyntäminen voi olla haasteellista.

Life cycle analysis and life cyle impact assessment methodologies: a state of the art

Life cycle analysis (LCA) is a comprehensive method for assessing the environmental impact of a product or an activity over its entire life cycle. The purpose of conducting LCA studies varies from one application to another. Different applications use LCA for different purposes. In general, the main aim of using LCA is to reduce the environmental impact of products through guiding the decision making process towards more sustainable solutions. The most critical phase in an LCA study is the Life Cycle Impact Assessment (LCIA) where the life cycle inventory (LCI) results of the considered substances related to the study of a certain system are transformed into understandable impact categories that represent the impact on the environment. In this research work, a general structure clarifying the steps that shall be followed ir order to conduct an LCA study effectively is presented. These steps are based on the ISO 14040 standard framework. In addition, a survey is done on the most widely used LCIA methodologies. Recommendations about possible developments and suggetions for further research work regarding the use of LCA and LCIA methodologies are discussed as well.

Ingénieur pédagogique et démarche projet : facteurs clés de succès pour l'intégration des technologies dans la pratique enseignante?

Résumé: Les gouvernements des pays occidentaux ont dépensé des sommes importantes pour faciliter l'intégration des technologies de l'information et de la communication dans l'enseignement espérant trouver une solution économique à l'épineuse équation que l'on pourrait résumer par la célèbre formule " faire plus et mieux avec moins ". Cependant force est de constater que, malgré ces efforts et la très nette amélioration de la qualité de service des infrastructures, cet objectif est loin d'être atteint. Si nous pensons qu'il est illusoire d'attendre et d'espérer que la technologie peut et va, à elle seule, résoudre les problèmes de qualité de l'enseignement, nous croyons néanmoins qu'elle peut contribuer à améliorer les conditions d'apprentissage et participer de la réflexion pédagogique que tout enseignant devrait conduire avant de dispenser ses enseignements. Dans cette optique, et convaincu que la formation à distance offre des avantages non négligeables à condition de penser " autrement " l'enseignement, nous nous sommes intéressé à la problématique du développement de ce type d'applications qui se situent à la frontière entre les sciences didactiques, les sciences cognitives, et l'informatique. Ainsi, et afin de proposer une solution réaliste et simple permettant de faciliter le développement, la mise-à-jour, l'insertion et la pérennisation des applications de formation à distance, nous nous sommes impliqué dans des projets concrets. Au fil de notre expérience de terrain nous avons fait le constat que (i)la qualité des modules de formation flexible et à distance reste encore très décevante, entre autres parce que la valeur ajoutée que peut apporter l'utilisation des technologies n'est, à notre avis, pas suffisamment exploitée et que (ii)pour réussir tout projet doit, outre le fait d'apporter une réponse utile à un besoin réel, être conduit efficacement avec le soutien d'un " champion ". Dans l'idée de proposer une démarche de gestion de projet adaptée aux besoins de la formation flexible et à distance, nous nous sommes tout d'abord penché sur les caractéristiques de ce type de projet. Nous avons ensuite analysé les méthodologies de projet existantes dans l'espoir de pouvoir utiliser l'une, l'autre ou un panachage adéquat de celles qui seraient les plus proches de nos besoins. Nous avons ensuite, de manière empirique et par itérations successives, défini une démarche pragmatique de gestion de projet et contribué à l'élaboration de fiches d'aide à la décision facilitant sa mise en oeuvre. Nous décrivons certains de ses acteurs en insistant particulièrement sur l'ingénieur pédagogique que nous considérons comme l'un des facteurs clé de succès de notre démarche et dont la vocation est de l'orchestrer. Enfin, nous avons validé a posteriori notre démarche en revenant sur le déroulement de quatre projets de FFD auxquels nous avons participé et qui sont représentatifs des projets que l'on peut rencontrer dans le milieu universitaire. En conclusion nous pensons que la mise en oeuvre de notre démarche, accompagnée de la mise à disposition de fiches d'aide à la décision informatisées, constitue un atout important et devrait permettre notamment de mesurer plus aisément les impacts réels des technologies (i) sur l'évolution de la pratique des enseignants, (ii) sur l'organisation et (iii) sur la qualité de l'enseignement. Notre démarche peut aussi servir de tremplin à la mise en place d'une démarche qualité propre à la FFD. D'autres recherches liées à la réelle flexibilisation des apprentissages et aux apports des technologies pour les apprenants pourront alors être conduites sur la base de métriques qui restent à définir. Abstract: Western countries have spent substantial amount of monies to facilitate the integration of the Information and Communication Technologies (ICT) into Education hoping to find a solution to the touchy equation that can be summarized by the famous statement "do more and better with less". Despite these efforts, and notwithstanding the real improvements due to the undeniable betterment of the infrastructure and of the quality of service, this goal is far from reached. Although we think it illusive to expect technology, all by itself, to solve our economical and educational problems, we firmly take the view that it can greatly contribute not only to ameliorate learning conditions but participate to rethinking the pedagogical approach as well. Every member of our community could hence take advantage of this opportunity to reflect upon his or her strategy. In this framework, and convinced that integrating ICT into education opens a number of very interesting avenues provided we think teaching "out of the box", we got ourself interested in courseware development positioned at the intersection of didactics and pedagogical sciences, cognitive sciences and computing. Hence, and hoping to bring a realistic and simple solution that could help develop, update, integrate and sustain courseware we got involved in concrete projects. As ze gained field experience we noticed that (i)The quality of courseware is still disappointing, amongst others, because the added value that the technology can bring is not made the most of, as it could or should be and (ii)A project requires, besides bringing a useful answer to a real problem, to be efficiently managed and be "championed". Having in mind to propose a pragmatic and practical project management approach we first looked into open and distance learning characteristics. We then analyzed existing methodologies in the hope of being able to utilize one or the other or a combination to best fit our needs. In an empiric manner and proceeding by successive iterations and refinements, we defined a simple methodology and contributed to build descriptive "cards" attached to each of its phases to help decision making. We describe the different actors involved in the process insisting specifically on the pedagogical engineer, viewed as an orchestra conductor, whom we consider to be critical to ensure the success of our approach. Last but not least, we have validated a posteriori our methodology by reviewing four of the projects we participated to and that we think emblematic of the university reality. We believe that the implementation of our methodology, along with the availability of computerized cards to help project managers to take decisions, could constitute a great asset and contribute to measure the technologies' real impacts on (i) the evolution of teaching practices (ii) the organization and (iii) the quality of pedagogical approaches. Our methodology could hence be of use to help put in place an open and distance learning quality assessment. Research on the impact of technologies to learning adaptability and flexibilization could rely on adequate metrics.

Coscheduling techniques and monitoring tools for non-dedicated cluster computing

Our efforts are directed towards the understanding of the coscheduling mechanism in a NOW system when a parallel job is executed jointly with local workloads, balancing parallel performance against the local interactive response. Explicit and implicit coscheduling techniques in a PVM-Linux NOW (or cluster) have been implemented. Furthermore, dynamic coscheduling remains an open question when parallel jobs are executed in a non-dedicated Cluster. A basis model for dynamic coscheduling in Cluster systems is presented in this paper. Also, one dynamic coscheduling algorithm for this model is proposed. The applicability of this algorithm has been proved and its performance analyzed by simulation. Finally, a new tool (named Monito) for monitoring the different queues of messages in such an environments is presented. The main aim of implementing this facility is to provide a mean of capturing the bottlenecks and overheads of the communication system in a PVM-Linux cluster.

Teaching fundamentals of computing theory: a constructivist approach

A Fundamentals of Computing Theory course involves different topics that are core to the Computer Science curricula and whose level of abstraction makes them difficult both to teach and to learn. Such difficulty stems from the complexity of the abstract notions involved and the required mathematical background. Surveys conducted among our students showed that many of them were applying some theoretical concepts mechanically rather than developing significant learning. This paper shows a number of didactic strategies that we introduced in the Fundamentals of Computing Theory curricula to cope with the above problem. The proposed strategies were based on a stronger use of technology and a constructivist approach. The final goal was to promote more significant learning of the course topics.

Development and application of computational methodologies in qualitative modeling

La présente thèse s'intitule "Développent et Application des Méthodologies Computationnelles pour la Modélisation Qualitative". Elle comprend tous les différents projets que j'ai entrepris en tant que doctorante. Plutôt qu'une mise en oeuvre systématique d'un cadre défini a priori, cette thèse devrait être considérée comme une exploration des méthodes qui peuvent nous aider à déduire le plan de processus regulatoires et de signalisation. Cette exploration a été mue par des questions biologiques concrètes, plutôt que par des investigations théoriques. Bien que tous les projets aient inclus des systèmes divergents (réseaux régulateurs de gènes du cycle cellulaire, réseaux de signalisation de cellules pulmonaires) ainsi que des organismes (levure à fission, levure bourgeonnante, rat, humain), nos objectifs étaient complémentaires et cohérents. Le projet principal de la thèse est la modélisation du réseau de l'initiation de septation (SIN) du S.pombe. La cytokinèse dans la levure à fission est contrôlée par le SIN, un réseau signalant de protéines kinases qui utilise le corps à pôle-fuseau comme échafaudage. Afin de décrire le comportement qualitatif du système et prédire des comportements mutants inconnus, nous avons décidé d'adopter l'approche de la modélisation booléenne. Dans cette thèse, nous présentons la construction d'un modèle booléen étendu du SIN, comprenant la plupart des composantes et des régulateurs du SIN en tant que noeuds individuels et testable expérimentalement. Ce modèle utilise des niveaux d'activité du CDK comme noeuds de contrôle pour la simulation d'évènements du SIN à différents stades du cycle cellulaire. Ce modèle a été optimisé en utilisant des expériences d'un seul "knock-out" avec des effets phénotypiques connus comme set d'entraînement. Il a permis de prédire correctement un set d'évaluation de "knock-out" doubles. De plus, le modèle a fait des prédictions in silico qui ont été validées in vivo, permettant d'obtenir de nouvelles idées de la régulation et l'organisation hiérarchique du SIN. Un autre projet concernant le cycle cellulaire qui fait partie de cette thèse a été la construction d'un modèle qualitatif et minimal de la réciprocité des cyclines dans la S.cerevisiae. Les protéines Clb dans la levure bourgeonnante présentent une activation et une dégradation caractéristique et séquentielle durant le cycle cellulaire, qu'on appelle communément les vagues des Clbs. Cet évènement est coordonné avec la courbe d'activation inverse du Sic1, qui a un rôle inhibitoire dans le système. Pour l'identification des modèles qualitatifs minimaux qui peuvent expliquer ce phénomène, nous avons sélectionné des expériences bien définies et construit tous les modèles minimaux possibles qui, une fois simulés, reproduisent les résultats attendus. Les modèles ont été filtrés en utilisant des simulations ODE qualitatives et standardisées; seules celles qui reproduisaient le phénotype des vagues ont été gardées. L'ensemble des modèles minimaux peut être utilisé pour suggérer des relations regulatoires entre les molécules participant qui peuvent ensuite être testées expérimentalement. Enfin, durant mon doctorat, j'ai participé au SBV Improver Challenge. Le but était de déduire des réseaux spécifiques à des espèces (humain et rat) en utilisant des données de phosphoprotéines, d'expressions des gènes et des cytokines, ainsi qu'un réseau de référence, qui était mis à disposition comme donnée préalable. Notre solution pour ce concours a pris la troisième place. L'approche utilisée est expliquée en détail dans le dernier chapitre de la thèse. -- The present dissertation is entitled "Development and Application of Computational Methodologies in Qualitative Modeling". It encompasses the diverse projects that were undertaken during my time as a PhD student. Instead of a systematic implementation of a framework defined a priori, this thesis should be considered as an exploration of the methods that can help us infer the blueprint of regulatory and signaling processes. This exploration was driven by concrete biological questions, rather than theoretical investigation. Even though the projects involved divergent systems (gene regulatory networks of cell cycle, signaling networks in lung cells), as well as organisms (fission yeast, budding yeast, rat, human), our goals were complementary and coherent. The main project of the thesis is the modeling of the Septation Initiation Network (SIN) in S.pombe. Cytokinesis in fission yeast is controlled by the SIN, a protein kinase signaling network that uses the spindle pole body as scaffold. In order to describe the qualitative behavior of the system and predict unknown mutant behaviors we decided to adopt a Boolean modeling approach. In this thesis, we report the construction of an extended, Boolean model of the SIN, comprising most SIN components and regulators as individual, experimentally testable nodes. The model uses CDK activity levels as control nodes for the simulation of SIN related events in different stages of the cell cycle. The model was optimized using single knock-out experiments of known phenotypic effect as a training set, and was able to correctly predict a double knock-out test set. Moreover, the model has made in silico predictions that have been validated in vivo, providing new insights into the regulation and hierarchical organization of the SIN. Another cell cycle related project that is part of this thesis was to create a qualitative, minimal model of cyclin interplay in S.cerevisiae. CLB proteins in budding yeast present a characteristic, sequential activation and decay during the cell cycle, commonly referred to as Clb waves. This event is coordinated with the inverse activation curve of Sic1, which has an inhibitory role in the system. To generate minimal qualitative models that can explain this phenomenon, we selected well-defined experiments and constructed all possible minimal models that, when simulated, reproduce the expected results. The models were filtered using standardized qualitative ODE simulations; only the ones reproducing the wave-like phenotype were kept. The set of minimal models can be used to suggest regulatory relations among the participating molecules, which will subsequently be tested experimentally. Finally, during my PhD I participated in the SBV Improver Challenge. The goal was to infer species-specific (human and rat) networks, using phosphoprotein, gene expression and cytokine data and a reference network provided as prior knowledge. Our solution to the challenge was selected as in the final chapter of the thesis.