Journées Scientifiques SEPIA

"Autour du HPC, du Cloud, de l'IoT et de l'Energie, visions sur les défis à relever"

12/13 Novembre 2015

Jeudi 12 Novembre

9h30 : Accueil et café

9h45 - 10h30 : Thierry Monteil, LAAS-CNRS

Title: OM2M an Open autonomic and semantic platform for M2M

Machine-to-Machine (M2M) concept is one of the main features of Internet of Things (IoT). It promises to inter-connect billions of devices in near future covering various domains. However, M2M suffers from a high vertical fragmentation of current M2M markets and lack of standards. To bridge this gap, the European Telecommunications Standards Institute (ETSI) and OneM2M consortium released a set of specifications for a common M2M service platform. We will present the open source Eclipse OM2M project, a horizontal ETSI and OneM2M compliant M2M service platform. OM2M provides a RESTful API to enhance interoperability. It proposes a modular architecture running on top of an OSGi Equinox runtime, making it highly extensible via plugins. It enables multiple communication protocols binding, reuse of existing remote devices management mechanisms, and interworking with existing legacy devices. OM2M is coupled with an autonomic manager using different knowledge bases to add “smart” behavior. Several use cases will illustrate our research on IoT.

10h30 - 11h15 : Raymond Namyst, Professeur, Université de Bordeaux

Title: Programming heterogeneous, manycore machines: a runtime system's perspective Heterogeneous manycore parallel machines, mixing multicore CPUs with manycore accelerators provide an unprecedented amount of processing power per node. Dealing with such a large number of heterogeneous processing units -- providing a highly unbalanced computing power -- is one of the biggest challenge that developpers of HPC applications have to face. To Fully tap into the potential of these heterogeneous machines, pure offloading approaches, that consist in running an application on host cores while offloading part of the code on accelerators, are not sufficient. In this talk, I will go through the major software techniques that were specifically designed to harness heterogeneous architectures, focusing on runtime systems. I will discuss some of the most critical issues programmers have to consider to achieve portability of performance, and how programming languages may evolve to meet such as goal. Eventually, I will give some insights about the main challenges designers of programming environments will have to face in upcoming years.

11h15 - 12h : Laurent Lefèvre, CR INRIA Lyon

Title : Towards energy proportional clouds, data centers and networks : the holy grail of energy efficiency ? Clouds, data centers and networks have become key players in the digital society. They host more and more services (centralized, distributed, virtualized) with increasing quality of service, availability and access requirements. However, these large scale distributed systems consume substantial and increasing amounts of energy. Energy consumption is the primary limiting factor in the deployment of such infrastructures. Several approaches are explored to reduce the electrical consumption of infrastructures during the use phase. The main problem comes from the fact that numerous infrastructures exhibit energy consumption that is not proportional to their use. This talk will present various works and initiatives combining hardware and software in order to achieve a proportional consumption of energy.

14h30 : Georges Da Costa, IRIT

Evaluation et optimisation de performance énergétique des centres de calcul

Georges présentera son habilitation à diriger les recherches. Cette HDR vise à répondre à la question "Comment gérer efficacement un centre de calcul" en fournissant les outils théoriques mais aussi pratiques nécessaires. Les centres de calculs sont au coeur de la vie d'une partie croissante de la population sans pour autant être visible, de par la prédominance des services en lignes. Leur consommation électrique est donc un enjeu d'actualité primordial, et le sera encore plus dans le futur prévisible. Gérer efficacement permet donc d'optimiser la qualité de service tout en réduisant leur consommation électrique. Ces travaux montrent les différents outils nécessaires. Ceux liés à la mesure, autant techniquement que de par la définition des métriques. Ensuite sont explorés les méthodes de modélisation de tels problèmes, ainsi que les techniques de résolution directes ou approchées. L'étape suivante consiste à étudier différentes heuristiques permettant une résolution approchée mais rapide du problème de gestion de centre de calcul. Diverses validations, autant expérimentales que basées sur des simulateurs améliorés serviront de support à ces démonstrations. Une ouverture vers le futur, grâce aux "datacenter in a box" distribués, hétérogènes conclue ces travaux.

==== This habilitation will answer to the challenge "How to efficiently manage a datacenter ?" by providing theoretical and practical tools. Datacenters are at the center of the life of an increasing part of the world population while being mostly unknown. As online services are heavily using datacenters, their role, while transparent, is of the utmost importance. Their electricity consumption is then a key challenge, and will be even more in a near future. Managing efficiently leads to optimization of quality of service, but also to improving energy efficiency. This work show the required tools. The tools links to measures and monitoring, from a technical point of view, but also by the definition of metrics. Then, it explores methods to model such problems, along with direct or approximated resolution techniques. The next step studies several heuristics to reach solutions fast but with approximations. Several types of validation, from experimental to improved simulated ones will serve as support to validate this work. As an opening toward the future, "datacenters in a box" will illustrate the highly distributed and heterogeneous trend of datacenters.

Vendredi 13 Novembre

10h : Patricia Stolf, IRIT

Patricia présentera son habilitation à diriger les recherches. Cette HDR présente ses travaux dans le contexte des systèmes informatiques à grande échelle pouvant être des grilles de calcul ou le cloud computing. Partant de deux constats : la consommation énergétique de ces systèmes est trop importante et ces systèmes sont de plus en complexes, ce mémoire se propose de répondre à la problématique suivante : comment gérer de manière optimale les ressources afin d'obtenir des infrastructures matérielles et logicielles "vertes" c'est à dire efficaces en énergie ? Ces travaux proposent trois axes de recherche : le premier en considérant le système complet et les leviers verts associés, le deuxième en étudiant des politiques d'allocation de ressources avec contraintes d'énergie et de chaleur, le troisième en étudiant des reconfigurations autonomiques d'applications. Pour finir, une description d'un centre de décision autonome pour des infrastructures vertes est proposée.

==== This habilitation will present my works in the context of the large-scale computer systems like grid or cloud computing. Based on two observations: the energy consumption of these systems is too important and these systems are getting more and complex to manage, this report suggests answering the following problem: how to optimize the resources use to obtain a "green" infrastructure and to improve energy efficiency? These works propose three areas of research: the first one by considering the complete system and the associated green leverages, the second by studying heuristics for energy and thermal aware placement, the third by using autonomic computing solutions to provide green reconfigurations. To finish, an autonomous and green decision center will be described.