The thesis consists of three parts. In the first part, we introduce the notion of device-cache-aware schedulers. Modern disk subsystems have many megabytes of memory for various purposes such as prefetching and caching. Current disk scheduling algorithms make decisions oblivious of the underlying device cache algorithms. In this thesis, we propose a scheduler architecture that is aware of underlying device cache. We also describe how the underlying device cache parameters can be automatically deduced and incorporated into the scheduling algorithm. In this thesis, we have only considered adaptive caching algorithms as modern high end disk subsystems are by default configured to use such algorithms. We implemented a prototype for Linux anticipatory scheduler, where we observed, compared with the anticipatory scheduler, upto 3 times improvement in query execution times with Benchw benchmark and upto 10 percent improvement with Postmark benchmark. The second part deals with implementing cooperative caching for the Redhat Global File System. The Redhat Global File System (GFS) is a clustered shared disk file system. The coordination between multiple accesses is through a lock manager. On a read, a lock on the inode is acquired in shared mode and the data is read from the disk. For a write, an exclusive lock on the inode is acquired and data is written to the disk; this requires all nodes holding the lock to write their dirty buffers/pages to disk and invalidate all the related buffers/pages. A DLM (Distributed Lock Manager) is a module that implements the functions of a lock manager. GFS’s DLM has some support for range locks, although it is not being used by GFS. While it is clear that a data sourced from a memory copy is likely to have lower latency, GFS currently reads from the shared disk after acquiring a lock (just as in other designs such as IBM’s GPFS) rather than from remote memory that just recently had the correct contents. The difficulties are mainly due to the circular relationships that can result between GFS and the generic DLM architecture while integrating DLM locking framework with cooperative caching. For example, the page/buffer cache should be accessible from DLM and yet DLM’s generality has to be preserved. The symmetric nature of DLM (including the SMP concurrency model) makes it even more difficult to understand and integrate cooperative caching into it (note that GPFS has an asymmetrical design). In this thesis, we describe the design of a cooperative caching scheme in GFS. To make it more effective, we also have introduced changes to the locking protocol and DLM to handle range locks more efficiently. Experiments with micro benchmarks on our prototype implementation reveal that, reading from a remote node over gigabit Ethernet can be upto 8 times faster than reading from a enterprise class SCSI disk for random disk reads. Our contributions are an integrated design for cooperative caching and lock manager for GFS, devising a novel method to do interval searches and determining when sequential reads from a remote memory perform better than sequential reads from a disk. The third part deals with selecting a primary network partition in a clustered shared disk system, when node/network failures occur. Clustered shared disk file systems like GFS, GPFS use methods that can fail in case of multiple network partitions and also in case of a 2 node cluster. In this thesis, we give an algorithm for fault-tolerant proactive leader election in asynchronous shared memory systems, and later its formal verification. Roughly speaking, a leader election algorithm is proactive if it can tolerate failure of nodes even after a leader is elected, and (stable) leader election happens periodically. This is needed in systems where a leader is required after every failure to ensure the availability of the system and there might be no explicit events such as messages in the (shared memory) system. Previous algorithms like DiskPaxos are not proactive. In our model, individual nodes can fail and reincarnate at any point in time. Each node has a counter which is incremented every period, which is same across all the nodes (modulo a maximum drift). Different nodes can be in different epochs at the same time. Our algorithm ensures that per epoch there can be at most one leader. So if the counter values of some set of nodes match, then there can be at most one leader among them. If the nodes satisfy certain timeliness constraints, then the leader for the epoch with highest counter also becomes the leader for the next epoch (stable property). Our algorithm uses shared memory proportional to the number of processes, the best possible. We also show how our protocol can be used in clustered shared disk systems to select a primary network partition. We have used the state machine approach to represent our protocol in Isabelle HOL logic system and have proved the safety property of the protocol.

Les interneurones de la couche moléculaire du cervelet (ICM: cellules en panier et cellules étoilées) sont connectés par des synapses électriques fréquentes et puissantes chez les jeunes rats et souris autour de la fin de la deuxième semaine postnatale. Les courants capacitifs des ICM montrent une composante lente qui reflète la charge des interneurones couplés électriquement. Leur analyse permet de quantifier le nombre de cellules directement couplées à une cellule et le nombre équivalent de cellules couplées (Alcami et Marty, soumis), et d'établir une difference de couplage entre les cellules en panier et les cellules étoilées pendant le développement postnatal. Elle a mené à proposer une topologie de réseau des cellules en panier. La force du couplage peut être modulée par les courants intrinsèques, dont Ih dans le domaine hyperpolarisant. Les synapses électriques modifient la propagation et les patrons d'activité dans le réseau des ICM en réponse à une excitation du réseau.L'étude de la connectivité des ICM par des synapses chimiques GABAergiques nous a mené à réexaminer les sources d'erreur des mesures d'activité électrique en configuration cellule attachée (Alcami et coll., 2012). Les mesures en cellule attachée peuvent modifier l'activité électrique des ICM en introduisant un couplage conductif entre la pipette d'enregistrement et l'intérieur cellulaire, résultant d'une combinaison de mécanismes de couplage passifs et actifs. / Molecular layer interneurons of the cerebellum (MLIs: basket cells and stellate cells) are connected by frequent and strong electrical synapses in young rats and mice around the end of the second postnatal week. Capacitive currents of MLIs show a slow component that reflects the charge of electrically-coupled MLIs. The analysis of capacitive currents makes it possible to quantify the number of directly connected cells and the equivalent number of coupled cells (Alcami and Marty, submitted). They were used to show a difference in coupling between basket and stellate cells and propose a model of the basket cell coupled network. Electrical coupling strength can be modulated by intrinsic currents, like the h current in the hyperpolarizing range. Electrical synapses modify the propagation and the patterns of activity in the MLI network, when the network is excited.The study of connectivity of MLIs by chemical GABAergic synapses led us to reevaluate the sources of error of cell-attached recordings (Alcami et al., 2012). Cell-attached measurements can modify cellular electrical activity of MLIs, by introducing a conductif coupling between the recording pipette and the cell interior, resulting from a combination of passive and active coupling.

La compréhension des mécanismes pilotes de la stabilisation des flammes-jet non-prémélangées constitue un point clé dans la caractérisation des modes opératoires des brûleurs industriels fonctionnant en régime de combustion diluée. Ce travail porte son attention sur l'étude expérimentale de l'influence de la dilution du combustible ou de l'air, sur le processus de décrochage et l'émission des polluants d'une flamme-jet non-prémélangée accrochée au brûleur. L'investigation est menée via un grand nombre d'expériences par combinaison des conditions suivantes : i) dioxyde de carbone (CO2), azote (N2), argon (Ar) et vapeur d'eau (H2Ov), sont utilisés comme diluants ; ii) deux configurations de dilution : dilution de l'air ou dilution du combustible ; iii) un couple de vitesses d'air et de combustible couvrant le domaine d'hystérésis de la flamme dans sa totalité, du régime de jet laminaire à celui de jet turbulent. Ceci permet de discriminer l'influence des effets intrinsèques à la nature du diluant de celle de l'aérodynamique des réactants (combustible et oxydant), dans la stabilité de la flamme accrochée. En particulier, les différences comportementales de la réponse de la flamme à la dilution de l'air ou à celle du combustible, sont analysées. Ces deux configurations de dilution diffèrent par deux effets de mélange, indépendants de la réaction, qui jouent un rôle important dans le cas de la dilution du combustible, mais sont négligeables dans le cas de celle de l'air : i) un effet dû à la modification de la fraction de mélange stœchiométrique. ii) un impact mécanique induit par l'apport de matière (diluants) responsable d'une augmentation de la vitesse des réactants. L'étude se divise en trois principales étapes. D'abord la réponse globale de la flamme à la dilution est étudiée via ses limites de décrochage quantifiées par les fractions molaires critiques des diluants dans l'oxydant ou dans le combustible, mesurées au décrochage. Le nombre de Peclet du combustible, Pef, est identifié comme le nombre adimensionnel qui ordonne ces limites de décrochage de manière homothétique pour tous les diluants. Grâce au comportement homothétique deux coefficients d'affinité, Kd,ox pour le cas de la dilution de l'air et Kd,f pour celle du combustible, sont introduits. Ils sont définis comme le rapport entre la limite de décrochage obtenue avec un diluant et celle obtenue avec le CO2 , à Pef = cste. Ceux-ci permettent l'établissement de deux polynômes génériques décrivant les limites de décrochage pour tous les diluants testés et dans toute la gamme des conditions aérodynamiques étudiées. En effet, Kd,ox et Kd,f englobent l'ensemble des effets physico-chimiques d'un diluant (dilution pure, thermique, propriétés de transport, chimie) et ceux des impacts mécaniques, affectant la stabilité de la flamme. Ils permettent de trouver les lois d'auto-similitude au décrochage pour un diluant chimiquement faible quelconque, à partir des résultats obtenus dans ce travail. Ensuite, une étude locale et détaillée du processus de décrochage induit par la dilution est réalisée. Celui-ci se base sur l'approche du bout propagatif décrivant la stabilité de la flamme accrochée comme résultant d'un équilibre à sa base entre la vitesse de l'écoulement et la vitesse de propagation. Afin de démontrer le lien entre cette approche et la stabilité de la flamme, une analyse approfondie des caractéristiques de sa base (localisation, intensité du radical CH* et champ de vitesses) est réalisée. Les résultats confirment la pertinence de l'approche du bout propagatif, comme mécanisme descriptif de la stabilisation de la flamme accrochée en présence de dilution. Enfin, une étude caractérisant aussi bien l'influence de la nature des diluants que celle de la configuration de dilution choisie (air ou combustible), sur l'émission des polluants (suies, NOx et CO), est présentée. / Understanding the main mechanisms piloting non-premixed jet flame stability is an important point in characterizing the operation modes of industrials burners in which dilution is involved. This work puts special emphasis on the experimental study of the influence of air-side and methane-side dilution in the lifting process of attached non-premixed jet flames. The study is based on numerous experiments combining the following conditions : i) carbon dioxide (CO2), nitrogen (N2), argon (Ar) or water vapor (H20v,) used as diluents d ; ii) two diluted configurations : air-side or methane-side dilution ; iii) two air and fuel velocities covering the entire flame hysteresis domain, from the laminar to the turbulent regime. This allows the influence of the intrinsic diluent nature effects to be discriminated from those of the aerodynamics of the reactants (fuel and oxidant), in attached flame stability. In particular, the behavioral differences of the flame response to air-side or to fuel-side dilution are analyzed. These two configurations differ by two mixing effects which are independent of the combustion reaction, and which are significant when the fuel is diluted, but negligible when air is diluted : i) an effect due to the changes in the stoichiometric mixture fraction ; ii) a mechanical impact induced by the addition of matter (diluents) producing an increase in the bulk velocity of the reactants. The study is composed of three parts. First, the global flame response to dilution is analyzed on the basis of the lifting limits defined as the critical molar fractions of the diluents in the fuel or in the oxidant measured at liftoff. The fuel Peclet number, Pef, appears as the dimensionless number which puts these limits in a homothetic order. This homothetic behavior allows the introduction of two affinity parameters, Kd,ox for air-side dilution and Kd,f for fuel-side dilution. They are defined by the ratio of the flame lifting limits calculated with a diluent d and with CO2, at Pef=const. Kd,ox and Kd, allow two generic polynomial laws to be established describing the flame lifting limits for all the diluents and in the whole range of aerodynamic conditions of this study. Indeed, Kd,ox and Kd,f encompass all the diluent effects affecting flame stability (pure dilution, thermal, transport, chemical), to which mechanical impacts are added. These coefficients make it possible to obtain the self-similarity laws of the lifting limits for any chemically-weak diluent, by using the results obtained in this work. Then, a local and detailed study of the flame lifting process induced by dilution is presented. This is based on the flame-leading-edge approach describing flame stability as a result of the balance between the incoming gas velocity of the reactants and the flame propagation velocity at the flame base. In order to show the link between this approach and flame stability, an extensive analysis of the flame-base characteristics (location, CH* emission intensity and velocity field) is carried out. The results attest to the pertinence of the propagative flame-leading-edge, as the mechanism describing the attached flame stability under dilution. Finally, a study concerning the influence of both the diluent nature and the diluted configuration (air or fuel) on pollutant emissions (soot, NOx and CO) is presented.