Filters








1,305 Hits in 2.8 sec

L'iris sauvage

Louise Glück, Marie Olivier
2014 Po&sie  
Enfin, la poétique de Glück explore l'intimité d'un sujet tout en éludant le personnel.  ...  avant l'attaque imminente, s'apprêtant à vaincre ces faiblesses, à ne succomber à rien, l'instant juste avant la floraison, l'ère de la maîtrise avant l'apparition du don, avant la possession. par Marie Olivier  ... 
doi:10.3917/poesi.149.0046 fatcat:ebhmf3ttord2dbymtrwrritsxu

The Boustrophedonic Turns Of Suffering in Faithful And Virtuous Night by Louise Glück

Marie Olivier
2021 Transatlantica : Revue d'Études Américaines  
The Boustrophedonic Turns Of Suffering in Faithful And Virtuous Night by Louise Glück Marie Olivier 1 In her latest collection, Faithful and Virtuous Night, published in 2014, Louise Glück explores a new  ...  (Glück, 2014 16, l. 190-193) 16 The lyrical instance is introduced as the very crux of the chiasmus where the outside and the inside meet, collide and converge.  ...  And yet, my silence was never total- (Glück, 2014 27, l. NOTES 1.  ... 
doi:10.4000/transatlantica.17145 fatcat:f4m2xhq6nrbfhbcevlshznx6xq

(D')Après Pouchkine : le jeu de la traduction dans « Omens » de Louise Glück

Marie Olivier
2013 Transatlantica : Revue d'Études Américaines  
Dans le cas de Louise Glück, Feinstein explique que la démarche préparatoire au travail fut la suivante : « Louise Glück [was] offered transliterations, stress patterns and word-for-word literals by Antony  ...  En traduisant ce poème de Pouchkine, Glück s'inscrit elle aussi dans la communauté plutôt large de ce « we poets ». 11 Dans « Education of the Poet », Glück évoque ses inspirations littéraires, et dit  ... 
doi:10.4000/transatlantica.6675 fatcat:pceha4pnwbchnpps6wfuuxks24

The Summer Garden Photograph, Or a Barthesian Reading of the Mutable Poetics of Louise Glück

Marie Olivier
2017 Journal of Computer Virology and Hacking Techniques  
The Summer Garden Photograph, Or a Barthesian Reading of the Mutable Poetics of Louise Glück Marie Olivier June 9, 1978 (Mourning) Not continuous, but immobile.  ...  Roland, American poetry Mots-clés: Glück Louise, deuil, photographie, punctum, diffèrement, Barthes Roland, poésie américaine Marie Olivier is an associate professor in American literature at Université  ... 
doi:10.4000/angles.1516 fatcat:v4x2nh4bpnftneyhet3bewfx5i

« De quoi ça parle »

Marie Olivier
2021 Transatlantica : Revue d'Études Américaines  
« De quoi ça parle » Hommage à Louise Glück, Prix Nobel de littérature 2020 Marie Olivier 1 Lors des différents entretiens journalistiques qui m'ont été demandés suite au Prix Nobel décerné à Louise  ...  Depuis l'enfance, Glück a toujours affirmé être attirée par les poèmes dont la voix est solitaire, et abhorrer la grandiloquence de poètes qui s'écouteraient parler (Glück 2020) : I was drawn, then as  ... 
doi:10.4000/transatlantica.16688 fatcat:zdmjvhylfvgjpj5beqr5nbs2ne

Improving the energy efficiency of software-defined backbone networks

Radu Carpa, Olivier Gluck, Laurent Lefevre, Jean-Christophe Mignot
2015 Photonic network communications  
Improving the Energy Efficiency of Software Defined Backbone Networks Radu CARPA, Olivier GLUCK, Laurent LEFEVRE, Jean-Christophe MIGNOT Inria Avalon -LIP Laboratory -cole Normale Suprieure of Lyon, University  ... 
doi:10.1007/s11107-015-0552-9 fatcat:76sxnywbj5gpjgseef74oh3msu

Splitting TCP for MPI Applications Executed on Grids

Olivier Glück, Jean-Christophe Mignot
2011 2011 IEEE Ninth International Symposium on Parallel and Distributed Processing with Applications  
In this paper, we first study the interaction between MPI applications and TCP on grids. Then, we propose MPI5000, a transparent applicative layer between MPI and TCP, using proxies to improve the execution of MPI applications on grids. Proxies aim at splitting TCP connections in order to detect losses faster and avoid to return in a slowstart phase after an idle time. Finally, we evaluate our layer executing the NAS Parallel Benchmarks on Grid'5000, the French research grid. The results show
more » ... at our architecture reduces the number of idle timeout and of long-distance retransmissions for BT, SP and LU benchmarks. Using MPI5000, these applications can decrease their execution time by 35%, 28%, and, 15% respectively. A comparison with MPICH-G2 performances shows that our layer can even outperform a grid enabled MPI implementation.
doi:10.1109/ispa.2011.11 dblp:conf/ispa/GluckM11 fatcat:2wlia4j6drb3vg6yhochatkxxe

An Efficient Network API for in-Kernel Applications in Clusters

Brice Goglin, Olivier Gluck, Pascale Vicat-Blanc Primet
2005 Proceedings IEEE International Conference on Cluster Computing  
Running parallel applications on clusters with high-speed local networks requires fast communication between computing nodes but also low latency and high bandwidth file access. However, the application programming interfaces of high-speed local networks were designed for MPI communication and do not always meet the requirements of other applications like distributed file systems. In this paper, we explore several solutions to improve the use of highspeed network for in-kernel applications.
more » ... ributed file systems implemented on top of the GM interface of MYRINET are first examined to demonstrate how hard it is to get an efficient interaction between such applications and the network. Then, we propose solutions to simplify and improve this interaction and integrate them into the kernel interface of the new MYRINET driver, MX. Performance comparisons between MX and GM, and their usage in both a distributed file system and a zero-copy protocol show nice improvements. Moreover, we are able to improve the performance of the flexible kernel API we designed in MX that allows to remove some intermediate copy. Résumé 1 Introduction The emergence of parallel applications led to the success of workstation clusters which are generic, extensible and less expensive. As the application always require more computational power, high-bandwidth and low-latency local networks with intelligent interface cards have been developed, such as MYRINET [BCF + 95], QUADRICS [PFHC03] or INFINI-BAND [Pfi01]. Specific software optimizations have been proposed to enable on one hand the overlapping of communications with computation phases, and on the other hand a drastic reduction of communication overhead in the host. Overlapping communication between different nodes is enabled by deporting a large part of the protocol stack in the network interface card. This led to high-level software programming interfaces that are based on asynchronous communication primitives, such as MPI (Message Passing Interface [For94]). The application posts send or receive requests and gets notification of their completion later. This is very different from traditional network interfaces, especially the SOCKET interface to access ETHERNET networks through the TCP/IP stack where all communication primitives are blocking. The reduction of the communication overhead in the host is achieved by three mechanisms. Intermediate copies have been removed (0-copy), the operating system has to be avoided (OS-bypass) and data have to be transferred directly between applications and the network through DMA (Direct Memory Access) initiated by the NIC (Network Interface Card). Several works have targeted communication software layers to improve the performance of parallel applications. However, very few works have focussed the utilization of highspeed local networks in other contexts, especially for applications that are not implemented in user-space like MPI computations. For instance, storage requirements in clusters may benefit from an efficient usage of these networks. This includes either distributed file systems or Network Block Devices. Applications using the SOCKET interface may also be improved by using a zero-copy protocol on top of the native in-kernel API of a high-speed local network. All these kernel contexts have specific constraints that are very different from user-space communication requirements. This paper presents our study and propositions for improving the interaction between such in-kernel applications and the highly-specific network programming interface on MYRINET. Section 2 presents potential in-kernel applications that may be used on clusters and their issues when interacting with the network. Then, we expose, in Section 3, how we have modified the GM interface of MYRINET to allow an efficient usage of the the underlying network in distributed file system clients. We detail in Section 4 several ideas to improve the flexibility of network programming interfaces and how we integrated them in the new MYRINET/MX driver. Section 5 finally gives a first performance evaluation in both distributed file systems and zero-copy socket protocols. Parallel applications running on clusters often want to get as much performance for storage access as for communication between computing nodes. Storage access layers may include either distributed file systems or network block devices, and were usually implemented in the operating system. The standard interface that is exposed to user application is basically Example of Memory Registration with GM on Myrinet networks DMA Memory User Kernel Network Address Translation (a) DMA transfer between the application and the network after memory registration.
doi:10.1109/clustr.2005.347044 dblp:conf/cluster/GoglinGP05 fatcat:54akmz4hprgqfo4iatmsinmtma

Segment routing based traffic engineering for energy efficient backbone networks

Radu Carpa, Olivier Gluck, Laurent Lefevre
2014 2014 IEEE International Conference on Advanced Networks and Telecommuncations Systems (ANTS)  
Energy consumption has become a limiting factor for deploying large-scale distributed infrastructures. This work 1 seeks to improve the energy efficiency of backbone networks by providing an intra-domain Software Defined Network (SDN) approach to selectively turn off a subset of links. We propose the STREETE framework (SegmenT Routing based Energy Efficient Traffic Engineering) that dynamically adapts the number of powered-on links to the traffic load. The core of the solution relies on SPRING,
more » ... a novel protocol being standardized by IETF. It is also known under the name of Segment Routing. The algorithms have been implemented and evaluated using the OMNET++ simulator. Experimental results show that the consumption of 44% of links can be reduced while preserving good quality of service.
doi:10.1109/ants.2014.7057272 dblp:conf/IEEEants/CarpaGL14 fatcat:s634nakjorgerhw6cwtmwrfq3m

Comparison and tuning of MPI implementations in a grid context

Ludovic Hablot, Olivier Gluck, Jean-Christophe Mignot, Stephane Genaud, Pascale Vicat-Blanc Primet
2007 2007 IEEE International Conference on Cluster Computing  
Today, clusters are often interconnected by long distance networks within grids to offer a huge number of available ressources to a range of users. MPI, the standard communication library used to write parallel applications, has been implemented for clusters. Two main features of grids: long distance networks and technological heterogeneity, raise the question of MPI efficiency in grids. This report presents an evaluation of four recent MPI implementations (MPICH2, MPICH-Madeleine, OpenMPI and
more » ... ridMPI) in the french research grid: Grid'5000. The comparison is based on the execution of pingpong, NAS Parallel Benchmarks and a real application in geophysics. We show that this implementations present performance differences. Executing MPI applications on the grid can be beneficial if the parameters are well tuned. The paper details the tuning required on each implementation to get the best performances. Résumé : Aujourd'hui les clusters sont souvent interconnectés par des réseaux longues distance à l'intérieur des grilles, dans le but d'offrir une plus grande capacité de calcul. MPI, la bibliothèque standard utilisée pour écrire des applications parallèles, a été concue pour les clusters. Cependant la grille apporte deux nouvelles contraintes : les réseaux longue distance et l'hétérogénéité. Celles-ci posent la question de l'efficacité de MPI dans ce contexte. Ce rapport présente une évaluation de quatre implémentations MPI récentes (MPICH2, MPICH-Madeleine, OpenMPI and GridMPI) dans la grille de recherche francaise Grid'5000. La comparaison est basée sur l'execution d'un pingpong, des NAS Parallel Benchmarks et d'une application réelle de géophysique. Nous montrons que les implémentations présentent des différences de performance. L'exécution d'applications MPI sur la grille peut être bénéfique si l'on effectue un paramètrage correct. Ce rapport présente également les optimisations nécessaires sur chaque implémentation MPI pour obtenir de meilleures performances. Mots-clés : MPI, implementations MPI, Grid'5000, paramètrage de TCP, optimisations de MPI Comparison and tuning of MPI implementations in grids
doi:10.1109/clustr.2007.4629265 dblp:conf/cluster/HablotGMGP07 fatcat:vahpit5mvzbfpl6neenmbxtgfm

Responsive algorithms for handling load surges and switching links on in green networks

Radu Carpa, Marcos Dias de Assuncao, Olivier Gluck, Laurent Lefevre, Jean-Christophe Mignot
2016 2016 IEEE International Conference on Communications (ICC)  
Responsive Algorithms for Handling Load Surges and Switching Links On in Green Networks Radu CÂRPA, Marcos Dias de ASSUNÇÃO, Olivier GLÜCK, Laurent LEFÈVRE , Jean-Christophe MIGNOT Inria Avalon -LIP Laboratory  ... 
doi:10.1109/icc.2016.7511557 dblp:conf/icc/CarpaAGLM16 fatcat:kuco7vt4urdulaiwi46temdm6m

Towards an energy estimator for fault tolerance protocols

Mohammed El Mehdi Diouri, Olivier Glück, Laurent Lefèvre, Franck Cappello
2013 Proceedings of the 18th ACM SIGPLAN symposium on Principles and practice of parallel programming - PPoPP '13  
Checkpointing protocols have different energy consumption depending on parameters like application features and platform characteristics. To select a protocol for a given execution, we propose an energy estimator that relies on an energy calibration of the considered platform and a user description of the execution settings.
doi:10.1145/2442516.2442561 dblp:conf/ppopp/DiouriGLC13 fatcat:fb3wo6yvxbhdxa5tfagggajb34

Designing and building SDN testbeds for energy-aware traffic engineering services

Marcos Dias de Assunção, Radu Carpa, Laurent Lefèvre, Olivier Glück, Piotr Boryło, Artur Lasoń, Andrzej Szymański, Michał Rzepka
2017 Photonic network communications  
As experimenting with energy-aware techniques on large-scale production infrastructure is prohibitive, a large number of proposed traffic-engineering strategies have been evaluated only using discrete-event simulations. The present work discusses (i) challenges towards building testbeds that allow researchers and practitioners to validate and evaluate the performance and quality of energy-aware traffic-engineering strategies, (ii) requirements to fulfill when porting simulations to testbeds,
more » ... (iii) two proof-of-concept testbeds. One testbed uses and provides Software-Defined Network (SDN) services created on the Open Network Operating System (ONOS) while the other is a composition of virtual Open vSwitches (OVS) controlled by the Ryu SDN framework. The aim of the testbeds is to validate previously proposed energy-aware traffic engineering strategies in different environments. We detail the platforms and illustrate how they have been used for performance evaluation. Additionally, the paper compares results obtained in the testbeds with evaluations performed using discrete-event simulations and presents challenges faced while implementing energyaware traffic engineering mechanisms as SDN services in testbed environments.
doi:10.1007/s11107-017-0709-9 fatcat:ykqmanicwzcdjalhikzlbcwnxm

Interaction efficace entre les réseaux rapides et le stockage distribué dans les grappes de calcul

Brice Goblin, Olivier Glück, Pascale Vicat-Blanc Primet
2008 Techniques et sciences informatiques  
Les applications parallèles s'exécutant sur les grappes nécessitent à la fois des communications performantes entre les différents noeuds et des accès efficaces au système de stockage. Nous proposons dans ce travail d'améliorer les performances du stockage distribué dans les grappes en utilisant au mieux le réseau haute performance sous-jacent pour accéder au stockage distant. Nous montrons que les besoins du stockage sont très différents de ceux du calcul parallèle et proposons différentes
more » ... tions pour résoudre, au niveau du système de fichiers, les problèmes liés au contrôle du réseau mais montrons qu'il est nécessaire de modifier l'interface de programmation réseau et le système d'explotation pour venir à bout des difficultés liées au transfert de données. Des expérimentations montrent qu'elles permettent une utilisation aisée et efficace des réseaux rapides dans le cadre du stockage distribué. Abstract: Parallel applications running on clusters require both high-performance communications between nodes and efficient access to the storage system. We propose to improve the performance of distributed storage systems in clusters by efficiently using the underlying high-performance network to access distant storage systems. We show that storage requirements are very different from those of parallel computation and that a modification of the network programming interface is required. We detail several propositions for these interfaces which make their utilization easier in the context of distributed storage. Performance evaluations show that their integration makes it easy to use and very efficient in the context of storage.
doi:10.3166/tsi.27.910-940 fatcat:dgkka75uu5catd2fopps474qoi

On Designing SDN Services for Energy-Aware Traffic Engineering [chapter]

Marcos Dias de Assunção, Radu Carpa, Laurent Lefèvre, Olivier Glück
2016 Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering  
As experimenting with energy-aware techniques on largescale production infrastructure is prohibitive, several traffic-engineering strategies have been evaluated using discrete-event simulation. The present work discusses (i) challenges towards building testbeds that allow researchers and practitioners to validate and evaluate the performance of energy-aware traffic-engineering strategies and (ii) requirements when porting simulations to testbeds. We discuss a proof-of-concept platform and an
more » ... lication that use and provide Software-Defined Network (SDN) services created on the Open Network Operating System (ONOS) to validate previously proposed energy-aware traffic engineering strategies. We detail the platform and illustrate how it has been used for performance evaluation.
doi:10.1007/978-3-319-49580-4_2 fatcat:n7jou27m2faofhpl3ac5vay2wq
« Previous Showing results 1 — 15 out of 1,305 results