Category: Events

CITI Seminar of Julie Dumas (Université Grenoble Alpes) on May 21 at 11am

Title: Cohérence de caches dans les architectures manycores et simulation

Date and Place: 21 / 05 / 2019 11:00 in TD-C

Host: Guillaume Salagnac

Abstract:

Les besoins en calcul toujours plus important ainsi que la prise en compte de l’efficacité énergétique ont conduit au développement des architectures manycores dont les travaux de recherche sont nombreux et en particulier autour du modèle mémoire. Si nous considérons les machines à mémoire partagée, le passage à l’échelle des protocoles de cohérence de caches est un problème encore ouvert. En effet, les protocoles basés sur l’espionnage, qui doivent transmettre à tous les caches les informations de cohérence, engendrent un nombre important de messages dont peu sont effectivement utiles. En revanche, les protocoles avec répertoires visent à n’envoyer des messages qu’aux caches qui en ont besoin. Dans ce cas, lorsque l’on augmente le nombre de cœurs, la taille du répertoire augmente en largeur et en profondeur et peut même dépasser la taille des données présentent dans les caches. Pour passer à l’échelle, un protocole doit émettre un nombre raisonnable de messages de cohérence et limiter le matériel utilisé pour la cohérence et en particulier pour la mémorisation du répertoire. Dans cette présentation, nous parlerons de DCC (Dynamic Coherent Cluster), une représentation dynamique de la liste des copies pour la cohérence de caches. Dans un second temps, nous nous intéresserons à la simulation de ces architectures qui freinent leur développement. En effet, plus un simulateur est précis plus les temps de simulation sont importants. Une des raisons est que la grande majorité des simulateurs ne s’exécutent pas en parallèle, ainsi une simulation de N cœurs est faite sur un seul cœur physique. Afin d’évaluer rapidement une architecture, nous proposerons un modèle de cache à haut niveau d’abstraction dans lequel des traces provenant d’un simulateur précis (gem5) sont injectées. DCC et d’autres représentations de la liste des copies seront évaluées à l’aide de cette méthodologie.

Biography:
Julie Dumas a obtenu sa thèse de doctorat en informatique à l’Université Grenoble Alpes en 2017. Ses travaux de recherche s’articulent autour de l’architecture des ordinateurs et plus particulièrement à la gestion de la mémoire dans les architectures de type manycores ainsi qu’aux techniques de simulation de ces dernières.


CITI Seminar of Thomas Begin (LIP, UCBL Lyon 1) on April 2 at 11am

Title: Contributions to the Performance Modeling of Computer Networks
Date and Place: 02 / 04 / 2019 11:00 in TD-C
Host: Jean-Marie Gorce and Florent de Dinechin
Abstract:
In this talk, I will present some of my contributions to the fields of performance evaluation and computer networks. I will first discuss a new modeling framework to evaluate the performance of DPDK-based virtual switches in the context of NFV (Network Function Virtualization) networks. Then, I’ll describe a scalable stochastic model to accurately forecast the performance of an IEEE 802.11-based network. Finally, I will introduce an original reduced-state description for multiserver queues that breaks the combinatorial complexity inherent to the classical state description and that can easily handle examples with hundreds of servers.

 

Biography:
Thomas Begin received his Ph.D. degree in Computer Science from UPMC (U. Paris 6) 2008. He was a post-doctoral fellow at UC Santa Cruz in 2009. Since 2009, he is an Associate Professor at UCBL (U. Lyon 1) in the Computer Science department. During the 2015-2016 academic year, he was on research leave at DIVA lab – University of Ottawa. T. Begin research interests are in performance evaluation, future network architecture, and system modeling. His principal applications pertain to high-level modeling, wireless networks, resource allocation and queueing systems.

CITI Seminar of Eddy Caron (LIP, École Normale Supérieure de Lyon) on March 19 at 11am

Title: Once upon a time … the deployment
Date and Place: 19/03/ 2019 11:00 in TD-C
Host: Jean-Marie Gorce and Florent de Dinechin
Abstract:
In large distributed systems the resource managements is one key of the efficient. And the deployment of the elements on resources are hidden everywhere, across the network, across the virtualization, across many infrastructures, etc. Through 6 stories we will discover many points of view of the deployment. First adventure, we will see how to deploy a middleware with self-stabilization skill. In the second story, be afraid, we will see how to deploy a secure Cloud Infrastructure. In the following story, we will introduce a deployment tool for reproducibility. The licenses deployment is another weird story with a lot of mysteries. An unbelievable story to deploy a data-driven microservices infrastructure. And finally, we will try to clear up the Fog deployment.

 

Biography:
Eddy Caron is an Associate Professor at Ecole Normale Supérieure de Lyon and holds a position with the LIP laboratory (ENS Lyon, France). He is a member of AVALON project from INRIA and Technical Manager for the DIET software package. He received his PhD in C.S. from University de Picardie Jules Verne in 2000 and his HDR (Habilitation à Diriger les Recherches) from the Ecole Normale Supérieure de Lyon in 2010. His research focuses on distributed computing environment, from P2P to Grid, Cloud and Edge computing. At the middleware level, he deals with a large scope of subjects (scheduling, workflow management, data management, energy management, security, software management, etc.)  with the same point of view of the resource magement in heterogeneous environments.
He is involved in many program committees (as HCW, IPDPS, ISPA, CloudTech, etc.). Since 2000, he contributed to more than 30 articles in journal or book chapter and more than 80 publications in international conferences. He was co-chair of the GridRPC working group in OGF. He was coordinator of two french ANR project (LEGO and SPADES). He was workpackage leader in the European project Seed4C around the security. He is the supervisor of 15 Phd (4 in progress). He teaches Distributed system, Architecture Operating System and Network, Grid and Cloud, etc. Moreover he was the Co-funder and Scientific Consultant of a company (SysFera). Deputy Director in charge of call for projects, research transfert and international affairs for the LIP Laboratory.  See http://graal.ens-lyon.fr/~ecaron for further information.

CITI Seminar of Alain Tchana (I3S, Université de Nice Sophia-Antipolis) on February 22 at 11AM

Title: Blablabla Virtualisation
Date and Place: 22/02/2019 at 11:00 in TD-C
Host: Jean-Marie Gorce and Florent de Dinechin
Abstract:
This talk will focus on virtualized infrastructure filed. In this domain, I aim at minimizing electricity consumption while improving application performance. To achieve the first goal, I work both at the entire datacenter level (by providing better VM placement strategies) and at the physical machine level (by providing better power management policies). Concerning the second goal, I work both at the VM monitor level (for minimizing its overhead) and at the VM’s operating system (OS) level (for making it aware of the fact that it is virtualized).

 

Biography:
Alain Tchana received his Ph.D. in computer science in 2011 at Institut National Polytechnique de Toulouse. The research topic of his Ph.D. was autonomic computing applied to cloud environments. He then spent two yeas as a postdoc at Université Joseph Fourier. During that time, he worked on building benchmarking systems. From September 2013 to September 2018, he was Associate Professor at Institut National Polytechnique de Toulouse. He was member of SEPIA research group at IRIT laboratory. His main research domain is virtualization. Since September 2018, he is full professor at Université de Nice Sophia-Antipolis. He is member of Scale research group at I3S. He continues to work in the virtualization domain.

CITI Talk: “Wired team presentation and discussions about blockchain”, Stéphane Frenot (INSA-Lyon, CITI) on February 15th at 11am

Title

Wired team presentation and discussions about blockchain

 

Summary

La blockchaines sont des technologies de stockage et de transmission d’informations, permettant la constitution de registres répliqués et distribués, sans organe central de contrôle, sécurisées grâce à la cryptographie, et structurées par des blocs liés les uns aux autres, à intervalles de temps réguliers. Elles sont utilisées par un certain nombre d’acteurs et suscitent de très nombreux débats aussi bien au coin café qu’à l’organisation mondiale du commerce.

En tant que membre du laboratoire de recherche CITI, je me sens concerné par ces technologies, et me demande de ce que nous devons en faire.
Je propose dans ce séminaire de vous présenter ma compréhension des systèmes blockchaines et de vous partager mon point de vue de concepteur d’applications distribuées et pair-à-pair sur le Web.

Bio

Stéphane Frénot a participé à la création du CITI en 2001. Il est spécialisé dans le génie logiciel et les application distribuées. Il a été responsable du thème middleware et de l’équipe INRIA Amazones au laboratoire. Puis il a participé au projet exploratoire INRIA Dice sur les plateformes d’intermédiations. Depuis 1an il est directeur du département Télécommunications Service et Usages de l’INSA où il enseigne le génie logiciel, les systèmes distribués et l’innovation.

Il a travaillé sur les architectures à composants logiciels, les systèmes pairs-à-pairs pour le déploiement de composants et sur un modèle de programmation orienté flux pour Javascript. Il a participé au dépôt de trois brevets : dans l’iot domestique, dans les flux javascript et dans un protocole de vote. Enfin il est responsable du développement de la plateforme Jumplyn de gestion de projets étudiants actuellement en test sur l’INSA pour la gestion des stages.


CITI Talk: “Wireless Networks Design in the Era of Deep Learning: Model-Based, AI-Based, or Both?”, Marco Di RENZO (CR CNRS, L2S) on February 13th at 11am

Title

Wireless Networks Design in the Era of Deep Learning: Model-Based, AI-Based, or Both?

 

Summary

This work addresses the use of emerging data-driven techniques based on deep learning and artificial neural networks in future wireless communication networks. In particular, a key point that will be made and supported throughout the work is that data-driven approaches should not replace traditional design techniques based on mathematical models. On the contrary, despite being seemingly mutually exclusive, there is much to be gained by merging data-driven and model-based approaches. To begin with, a detailed presentation is given for the reasons why deep learning based on artificial neural networks will be an indispensable tool for the design and operation of future wireless communications networks, as well as a description of the recent technological advances that make deep learning practically viable for wireless applications. Our vision of how artificial neural networks should be integrated into the architecture of future wireless communication networks is presented, explaining the main areas where deep learning provides a decisive advantage over traditional approaches. Afterwards, a thorough description of deep learning methodologies is provided, starting with presenting the general machine learning paradigm, followed by a more in-depth discussion about deep learning. Artificial neural networks are introduced as the peculiar feature that makes deep learning different and more performing than other machine learning techniques. The most widely-used artificial neural network architectures and their training methods will be analyzed in detail. Moreover, bridges will be drawn between deep learning and other major learning frameworks such as reinforcement learning and transfer learning. After introducing the deep learning framework, its application

to wireless communication is addressed. This part of the work first provides the state-of-the-art of deep learning for wireless communication networks, and then moves on to address several novel case-studies wherein the use of deep learning proves extremely useful for network design. In particular, the connection between deep learning and model-based approaches is emphasized, proposing several novel techniques for cross-fertilization between these two paradigms. For each case-study, it will be shown how the use of (even approximate) mathematical models can significantly reduce the amount of live data that needs to be acquired/measured to implement data-driven approaches. For each application, the merits of the proposed approaches will be demonstrated by a numerical analysis in which the implementation and training of the artificial neural network used to solve the problem is discussed. Finally, concluding remarks describe those that in our opinion are the major directions for future research in this field.


Seminar of Carles Antón-Haro (Telecommunications Technological Center of Catalonia) on September 27th at 9.30am

The next CITI seminar will take place on September 27th, at 9.30am in the margin of Yuqi PhD defense. This seminar entitled “Machine- and Deep-Learning for Beam Selection in Hybrid Analog Beamforming Architectures” will be presented by Carles Antón-Haro from Telecommunications Technological Center of Catalonia.

Titre : Machine- and Deep-Learning for Beam Selection in Hybrid Analog Beamforming Architectures

Abstract : This talk deals with the application of deep learning (DL) and machine learning (ML) techniques to beam selection problems in the uplink of a mmWave communication system. Specifically, we consider a hybrid beamforming architecture comprising an analog beamforming (ABF) network followed by a zero-forcing baseband processing block. The goal is to select the optimal configuration for the ABF network based on the estimated AoAs of the various user equipments. To that aim, we consider (i) two supervised machine-learning approaches: k-nearest neighbors (kNN) and support vector classifiers (SVC); and (ii) a feed-forward deep neural network: the multilayer perceptron (MLP). Computer simulations reveal that, for a well-designed codebook of analog beamformers, this task can be effectively accomplished by such data-driven schemes. Performance, in terms of sum-rate, is very close to that achievable via exhaustive search, in particular for the MLP.

Bio : Carles Antón-Haro received his PhD degree in Telecommunications Engineering from the Technical University of Catalonia in 1998 (cum-laude). He also holds a Master in Business Administration (MBA) from EADA Business School (2014, Best Final Degree Project Award). In the pursuit of his PhD degree, he was a recipient of scholarship granted by the Dept. of Universities and Research of the Generalitat de Catalunya (1995-1998). As a Research Assistant (1994-1998, UPC) and Research Associate (1998-1999, UPC), he participated in several EC Projects (Tsunami, Tsunami II, Sunbeam), as well as in other projects funded by the Spanish government. He was Teaching Assistant in the field of Computer Architecture (UPC 1994, UOC 1998-2012). In 1999, he joined Ericsson Spain, where he participated in two rollout projects of 2G and 3G mobile networks (2000, Regional Coordinator).

Currently, he is with the CTTC, where he works as a Director of R&D Programs and Senior Research Associate. Main duties, in addition to his daily research activity, include the overall coordination of CTTC’s participation in publicly-funded R&D projects and technology transfer projects, networking activities towards the establishment of strategic alliances with the industry and academia, monitoring of R&D funding programs and identification of new opportunities, and interaction with CTTC’s Scientific Committee in what concerns R&D programs. In the past, he was also in charge of the recruitment of human resources at the CTTC, and he defined the internal processes in relation with CTTC’s project portfolio. Since 2001, he has promoted or coordinated over 60 R&D projects or proposals and has been directly involved in some of them (P2P SmartTest, NEWCOM#, EXALTED, eCROPS, ADVANTAGE, WINNER, to name a few). He is an elected member of the Steering Board of the Networld2020 European Technology Platform since 2009 (formerly known as Net!works).

His research interests are in the field of signal processing for communications, this including radio interface design, multi-user MIMO, wireless sensor networks, opportunistic communications, link layer protocols (MAC, H-ARQ); and estimation theory with emphasis in state estimation for Smart Electricity Grids. He has published +20 technical papers in IEEE journals, books and book chapter; as well as +80 papers in international and national conferences. He is a recipient of the 2015 Best Paper Award of the Transmission, Access, and Optical Systems (TAOS) Technical Committee’s (Green Communications Track, ICC). He has supervised four Master Theses and five PhD Theses (two in progress). He has also acted as a reviewer of project proposals for various (inter)national funding agencies (e.g., ANEP, AGAUR, MIUR, ANR, ANVUR) and takes part in PhD Evaluation Committees on a regular basis.

 


Seminar of Lionel MOREL (CEA Grenoble) on October 18th at 2pm

The next CITI seminar will take place on October 18 th, at 2pm. This seminar entitled “Polen: une approche SW/HW pour la confidentialité des programmes et des données” will be presented by our former colleague Lionel MOREL from CEA Grenoble.

Titre : Polen: une approche SW/HW pour la confidentialité des programmes et des données.

Résumé : D’aucuns voudraient connecter un nombre grandissant d’objets, entre eux, mais aussi au réseau internet, pour permettre la collecte d’un nombre toujours plus grand d’informations et réaliser ainsi l’augmentation de nos vies jusque-là visiblement sous-dimensionnées. Mais connecter des objets, y stocker des informations personnelles, tout en les rendant accessibles facilement au reste du monde ouvre la voie à toute une série d’usages dangereux pour nos données et nous-mêmes.

Les approches de protection matérielles traditionnellement utilisées (eg pour la carte à puce) sont certes très efficaces, mais elles sont également très coûteuses en développement, en certification, et en déploiement. Au CEA, nous étudions comment des approches logicielles peuvent venir en complément de ces approches matérielles pour augmenter le niveau de confiance placé dans l’objet tout en limitant les coûts et en flexibilisant l’application des protections.

Dans cet exposé, je présenterai un cas particulier d’approche mêlant contre-mesures matérielles et logicielles, que nous développons actuellement. Il sera question d’attaques par canaux cachés, de reverse-engineering (un peu) de compilation dynamique, de chiffrement de code (plus), de pompe à insuline et de lampes connectées aussi, et de fin du monde peut-être.

Bio : Après une thèse sur les langages de programmation dédiés aux systèmes critiques, soutenue à Verimag en 2005 et quelques voyages scientifico-culturels en Bretagne et Finlande, Lionel Morel a intégré l’INSA Lyon en 2007 et le CITI en 2009. Il y a mené des travaux de recherche entre autres sur la programmation et l’évaluation de performances de machines parallèles, tout en enseignant les systèmes d’exploitations, l’architecture des ordinateurs et la compilation. Depuis 2017, il est détaché auprès du CEA, à Grenoble, où il travaille sur l’usage de la compilation pour la sécurité.