Avalon is contributing to LLVM OpenMP/runtime

Philippe Virouleau, funding thanks to EoCoE-II project, has proposed patches to LLVM OpenMP/runtime in order to provide better control and performances of OpenMP task execution. First accepted patch was pushed in LLVM master branch (https://reviews.llvm.org/D63196). It solves an side effect due to the task throttling heuristic that serializes task execution. It may  cripple the application performance in some specific task graph scenarios, like the ones detailed in section 4.2 from this paper published at IWOMP 2018 (the full text can be found here). In such cases not having the full task graph prevent some opportunities for cache reuse between successive tasks.

Mid term goals are to transfer some important and innovative features mostly already available in libKOMP https://gitlab.inria.fr/openmp/libkomp.

 

 

Avalon at the TILECS Workshop

Avalon at the TILECS Workshop

Towards an Infrastructure for Large-Scale Experimental Computer Science

3 et 4 juillet, Campus Grenoble, France

More information about SILECS and the TILECS Workshop: https://www.silecs.net/tilecs-2019/

Laurent Lefevre (Avalon / Speaker) and Georges Da Costa with Christian Perez (Avalon / Chairman)

SILECS is a research infrastructure being built to gather the efforts of several testbeds.

 

The Avalon team are involved in this impressive project.

Farah Ait Salaht, Eddy Caron, Marcos Dias de assuncao, Laurent Lefevre (speaker), Matthieu Imbert, Christian Perez  (Session Chairman) was at this workshop.

Journée STIP Inria – Rencontre Avalon

The STIP Inria day started

Journée Ateliers Valorisation organisé par le STIP (Service Transfert Innovation et Partenariat ) Inria le 18 juin au LIP.

Organisation: STIP Inria (Helena Jerôme, Philippe Broun)

Intervenants: Pascale Vicat Blanc, Olivier Arsene, Xavier Blot

Lieu: Salle de réunion du M7

Programme

9h-12h : ateliers d’introduction à l’entrepreneuriat, présentations des dispositifs Inria (accompagnement, financement…), témoignage et retour d’expérience et échanges avec Pascale Vicat Blanc, scientifique-entrepreneure Inria (spécialisée dans le Cloud Computing)
13h15-16h15 : Ateliers en groupe basés entre autre sur une méthode de créativité : choix d’une techno/sujet dans les sujets scientifiques de l’équipe Avalon et sur les domaines d’applications des résultats de recherches, pour illustrer l’atelier, construction d’un livrable utile aux participants (encore à définir)

Participants: Marouane Azzouz, Dorra Boughzala, Eddy Caron, Alexandre Da Silva Veith, Felipe De Souza, Zakaria Fraoui, Mohamed Hamnache, Zeina Houmani, Cyril Séguin.

The day in live


10h15 : The Marshmallow Challenge

Organized by Beyond Lab (Xavier Blot and Olivier Arsene)

18 minutes to build the Marshmallow Tower …

Alexandre, Zeina and Felipe Team

 

 

Cyril, Dorra and Zakaria Team

 

 

 

 

 

 

 

And the winner is …

Cyril, Dorra and Zakaria Team


10h35: What’s STIP Inria can do for you [Philippe Broun]

Philippe Broun’s Talk – What’s the STIP can do for you

If you are lost in Innovation world… Who ya gonna call?

 

 

 

 

 

 

 

 

 


10h57: Start’up experience: an amazing human adventure [Pascale Vicat-Blanc]

The Pascale Vicat-Blanc’s start-up experience

 

 

 

 

 

 

 

 


12h15: Beyond Lab introduction [Xavier Blot and Olivier Arsene]

The website: beyondlab.org

Beyond lab introduction


 

Lunch … discussions continue with the help of a good Pizza …

 

 

 

 

 

 

 


14h05-16h45 Workshop to introduce three Avalon’s projects

A game with three teams.

Stage 1: Elevator pitch

SeeDep, SkyData and FogMonitoring

Stage 2: Team, appropriate, go beyond… Find the value

Each team try to find the value of an other project.

Stage 3: Final Pitch and Strategy

 

Thesis Defense – Hadrien Croubois

Hadrien Croubois: Toward an autonomic engine for scientific workflows and elastic Cloud infrastructure

Everyone is welcome to attend Hadrien Croubois’s thesis defense, which will take place Tuesday 16th October at 14h at Salle des thèse (ENS de Lyon).

You are also invited to the cocktail that follows the defense.

Advisors:

Eddy Caron, ENS de Lyon


Committee members:

Noël De Palma, Université Jospeh Fourier, reviewer
Johan Montagnat, CNRS, Laboratoire I3S UMR 7271, reviewer
Luciana Arantes, Université Sorbonne, examiner
Frédéric Desprez, Inria, examiner
Pushpinder Kaur Chouhan,Ulster University, examiner


Abstract:

The constant development of scientific and industrial computation infrastructures requires the concurrent development of scheduling and deployment mechanisms to manage such infrastructures. Throughout the last decade, the emergence of the Cloud paradigm raised many hopes, but achieving full platform autonomicity is still an ongoing challenge.

Work undertaken during this Ph.D. aimed at building a workflow engine that integrated the logic needed to manage workflow execution and Cloud deployment on its own. More precisely, we focus on Cloud solutions with a dedicated Data as a Service (DaaS) data management component. Our objective was to automate the execution of workflows submitted by many users on elastic Cloud resources.

This contribution proposes a modular middleware infrastructure and details the implementation of the underlying modules:

  • A workflow clustering algorithm that optimises data locality in the context of DaaS-centered communications;
  • A dynamic scheduler that executes clustered workflows on Cloud resources;
  • A deployment manager that handles the allocation and deallocation of Cloud resources according to the workload characteristics and users’ requirements.

All these modules have been implemented in a simulator to analyse their behaviour and measure their effectiveness when running both synthetic and real scientific workflows. We also implemented these modules in the DIET middleware to give it new features and prove the versatility of this approach. Simulation running the WASABI workflow (waves analysis based inference, a framework for the reconstruction of gene regulatory networks) showed that our approach can decrease the deployment cost by up to 44% while meeting the required deadlines.

Phd Crossover

Phd Crossover

LIP & LBMC – Avalon & Gandrillon – Hadrien & Arnaud –

En octobre vous pouvez découvrir le crossover entre la thèse d’Hadrien Croubois (Avalon) du laboratoire de l’Informatique du Parallélisme (LIP) de l’ENS  et d’Arnaud Bonnafoux (Gandrillon) du Laboratoire de Biologie Moléculaire de la Cellule (LBMC) de l’ENS.


au LBMC

Commencez par découvrir le vendredi 12 octobre 2018, à 14h à l’ENS de Lyon site Monod (salle de réunion M6) comment Arnaud Bonnafoux a conçu WASABI. (Plus d’information ici)


au LIP

Puis venez découvrir le mardi 16 octobre 2018, à 14h à l’ENS de Lyon site Monod (salle des thèses) comment le moteur de workflow d’Hadrien Croubois est profitable aux besoins en calcul de WASABI. (Plus d’information ici)


Thesis Defense — Pedro Silva

Everyone is welcome to attend Pedro Silva’s thesis defense, which will take place next Monday 11th December at 14h at Salle Place de l’Ecole (1, Place de l’École, 69003).

Title:
On the mapping of distributed applications onto multiple Clouds.

Advisors:
Christian Perez, Inria
Frédéric Desprez, Inria

Committee members:
Patricia PASCAL-STOLF, IUT de Blagnac, reviewer
Emmanuel Jeannot, Inria, reviewer
Christine Morin, Inria, examiner
Erik Elmroth, University of Umea – Sweden, examiner

You are also invited to the cocktail that follows the defense at the
LIP’s coffee area (coin café) (46 Allée d’Italie, 3rd floor).

Abstract:

The Cloud has become a very popular platform for deploying distributed applications. Today, virtually any credit card holder can have access to Cloud services. There are many different ways of offering Cloud services to customers. In this thesis we specially focus on the Infrastructure as a Service (IaaS), a model that, usually, proposes virtualized computing resources to costumers in the form of virtual machines (VMs).
Thanks to its attractive pay-as-you-use cost model, it is easier for customers to outsource hosting infrastructures and benefit of
savings related to upfront investments and maintenance costs. Also, customers can have access to features such as scalability, availability, and reliability, which previously were almost exclusive for large companies.

To place a distributed application on the Cloud, a customer must first consider the mapping between her application (or its parts) to the target infrastructure. She needs to take into consideration cost, resource, and communication constraints to select the most suitable set of VMs, from private and public Cloud providers. However, defining a mapping manually may be a challenge in large-scale or time constrained scenarios since the number of possible configuration explodes. The large offer of Cloud providers in the market and eventual advantages of having an application deployed over different Cloud sites, like redundancy and reachability, for example, make this challenge even more complex. Furthermore, when automating the mapping definition, scalability issues must be taken into account since this problem is a generalization of the graph homomorphism problem, which is NP-complete.

In this thesis we address the problem of calculating initial and reconfiguration placements for distributed applications over possibly multiple Clouds. Our objective is to minimize renting and migration costs while satisfying applications’ resource and communication constraints. We concentrate on the mapping between applications and Cloud infrastructures. Using an incremental approach, we split the problem into three different parts and propose efficient heuristics that can compute good quality placements very quickly for small and large scenarios. First we model the problem as a communication oblivious problem and propose vector packing based heuristics able to calculate initial placement solutions. Then, we extend our application and infrastructure models by introducing communication constraints, and propose a graph homomorphism based heuristic to calculate initial placement solutions. In the last part, we introduce application reconfiguration to our models and propose a heuristic able to calculate communication and reconfiguration aware placement solutions.

These heuristics have been extensively evaluated against state of the art approaches such as MIP solvers and meta-heuristics. We show through simulations that the proposed heuristics manage to compute solutions in a few seconds that would take many hours or days for other approaches to compute.

3ème journée du thème IDCHP de la FIL

Troisième journée du thème Informatique distribuée et calcul haute performance (IDCHP) de la Fédération Informatique de Lyon

Le 13 octobre à 9h30 à L’ENS de Lyon. Salle des conseils de l’ENS. Batiment Monod (2ème étage)
Comment venir: Accès à l’ENS. Site de Monod.
Inscription (gratuite mais fortement souhaitée pour des questions de logistique via le pad communiqué dans la mailing list)

Journée d’échange axée sur les systèmes distribuées

&

Talk invité ouvert à tous: The Red Belly Blockchain: Consistency and Scalability. Vincent Gramoli (Université de Sydney).  (voir détails ci-dessous)

 Suite à la première journée du thème IDCHP de la FIL nous avons prévu de nous retrouver sur des thèmes plus spécifiques pour des journées d’échange.

  • La deuxième journée sur l’IoT et les capteur aura lieu le 27 novembre (lieu et détails à venir)
  • La troisième journée sur les systèmes distribuées le 13 octobre (voir le programme ci-dessous)

Programme:

9h30 – 9h45 Accueil

9h45 – 12h00 Session plénière

Quelques sujets que nous pouvons aborder …

Autonomic workflow Deployment on dynamic platforms Hadrien Croubois (LIP) / Eddy Caron (LIP)
Taming Tail Latency in Key-Value Stores for Heterogeneous Workloads Sonia Ben Mokhtar (LIRIS)
Blockchain: an Ephemeral Buzzword or a New Paradigm for Distributed Systems? – What Research Agenda for the Next 5 Years?  Lionel Brunie (LIRIS)

Achieving Privacy and Utility Trade-off in User Mobility Data Sara Bouchenak (LIRIS)
 

12h00 – 13h30 Repas offert par la FIL

14h00Exposé ouvert à tous à l’Amphi B – Site de Monod – ENS de Lyon

Titre: The Red Belly Blockchain: Consistency and Scalability
Vincent Gramoli (University Sydney)
Abstract: To cope with blockchain inconsistencies, like double-spending, developers started building upon Byzantine Fault-Tolerant (BFT) consensus to solve this blockchain problem. At first it seems reasonable because consensus can be effective at totally ordering transactions into a chain.
Unfortunately, these two problems are different. In short, the blockchain problem aims at totally ordering blocks of transactions issued by a large set of potentially misbehaving internet machines, whereas the consensus problem aims at deciding upon one of the values proposed by typically fewer but non-misbehaving machines.
In this talk, I will describe the blockchain consensus problem, a variant of the classic consensus definition, that was recently defined for blockchains. I will present the red belly blockchain, a blockchain that does not use BFT consensus as a black box but rather solves the blockchain consensus problem and illustrate empirically the benefits it offers.
Bio: Vincent Gramoli is the Head of the Concurrent Systems Research Group at the University of Sydney and a Senior Researcher at Data61-CSIRO.
He received his PhD from University of Rennes and his Habilitation from UPMC Sorbonne University. Prior to this, he was affiliated with Cornell University and EPFL. His research interest is in distributed computing.

Participants:

Ben Mokhtar Sonia LIRIS
Bennani Nadia LIRIS
Bouchenak Sara LIRIS
Brunie Lionel LIRI
Caniou Yves LIP
Caron Eddy LIP
Cazabet Rémy LIRIS
Croubois Hadrien LIP
Furno Angelo CITI
Gramoli Vincent Sydney University
Kotto Kombi Roland LIRIS
Le Mouël Frédéric CITI
Lefevre Laurent LIP
Moyaux Thierry DISP
Médini Lionel LIRIS
Perez Christian LIP
Simonin Olivier CITI
Urbain Xavier LIRIS
Vivien Frédéric LIP
Canon Louis-Claude LIP

 

 

 

 

 

 

Pour toutes questions merci de contacter Eddy Caron et/ou Sonia Ben Mokhtar.

Posted in FIL

GrPPI: A Generic Parallel Pattern Interface for Stream and Data Processing

Next Avalon working group will be tomorrow (3/10/2016) at 15h in amphi L. Dr. Manuel F. Dolz, currently visiting us from Madrid, will be talking about his work.

GrPPI: A Generic Parallel Pattern Interface for Stream and Data Processing

Current parallel programming frameworks aid developers to a great extent in implementing applications that exploit parallel hardware resources. Nevertheless, developers require additional expertise to properly use and tune them to operate efficiently on specific parallel platforms. With the lack of high-level parallel pattern abstractions, we present GrPPI, a generic and reusable parallel pattern interface for both stream processing and data-intensive C++ applications (https://github.com/arcosuc3m/grppi). GrPPI accommodates a layer between developers and existing parallel programming frameworks targeting multi-core processors, such as C++ threads, OpenMP and Intel TBB. To achieve this goal, this interface leverages modern C++ features, metaprogramming techniques, and template-based programming to act as switch among those frameworks. All in all, thanks to its high-level API and compact design, GrPPI allows users to easily expose parallelism and hide away the complexity behind concurrency mechanisms. We evaluate this interface using an image processing use case and demonstrate its benefits from the usability, flexibility, and performance points of view.