Section: Partnerships and Cooperations

National Initiatives

Participants : Sébastien Badia, Sylvain Contassot-Vivier, Stéphane Genaud, Jens Gustedt, Emmanuel Jeanvoine, Lucas Nussbaum, Martin Quinson, Tinaherinantenaina Rakotoarivelo, Luc Sarzyniec, Stéphane Vialle.

INRIA ADT SimGrid for human beings (2009–2011)

SimGrid for human beings is another INRIA Technological Development Action aiming at providing engineering manpower to the SimG rid project to improve the documentation and to provide stock implementations of classical algorithms in order to ease its usage by the users. Mehdi Fekari was hired on this project, leading to the results described in Section  6.2.1 .

INRIA ADT Aladdin-G5K (2007–2012?)

ADT Aladdin-G5K (A LArge-scale Distributed Deployable INfrastructure) is an INRIA Technological Development Action. It is a management structure for Grid'5000. The goal of Aladdin-G5K is to further develop the Grid'5000 testbed, and perform system administration and maintenance, as well as additional tasks such as maintaining the various tutorials. Three engineers from Nancy (two from AlGorille, one from SED) are contributing to Aladdin-G5K: Tina Rakotoarivelo (hired for the ADT), Sébastien Badia (IE CPER) and Benjamin Dexheimer (SED).

INRIA ADT Kadeploy (2011–2013)

ADT Kadeploy focuses on the Kadeploy software. Kadeploy is a tool for efficient, scalable and reliable cluster deployment, used on several clusters at INRIA and playing a key role on the Grid'5000 testbed. This ADT allows us to continue the development of Kadeploy, by improving its performance, its reliability and its security. We are also adding features that are required in some contexts so that it will be possible to further distribute Kadeploy and increase its usage. During the ADT, we are collaborating with INRIA platforms that do not use Kadeploy yet, but have similar needs, as well as with industry. Luc Sarzyniec was hired as young engineer (IJD) as part of this project.

INRIA ADT Solfege (2011–2013)

ADT Solfege (Services et Outils Logiciels Facilitant l'Experimentation á Grande Échelle) aims at developing or improving a tool suite for experimentation at large scale on testbeds such as Grid'5000. Specifically, we will work on control of a large number of nodes, on data management, and on changing experimental conditions with emulation. Emmanuel Jeanvoine (SED) is delegated in the AlGorille team for the duration of this project.

INRIA AEN HEMERA

Héméra is an INRIA Large Wingspan project, started in 2010, that aims at demonstrating ambitious up-scaling techniques for large scale distributed computing by carrying out several dimensioning experiments on the Grid'5000 infrastructure, at animating the scientific community around Grid'5000 and at enlarging the Grid'5000 community by helping newcomers to make use of Grid'5000.

Within that project, Martin Quinson, Lucas Nussbaum and Stéphane Genaud lead three working groups, respectively on simulating large-scale facilities, on conducting large and complex experimentations on real platforms, and on designing scientific applications for scalability.

CNRS initiatives, GDR-ASR and specific initiatives

We participate at numerous national initiatives. In the GDR-ASR (architecture, systems, and networks) we take part in RGE action (Réseau Grand Est). The finances of RGE, led by Stéphane Vialle at SUPÉLEC, are provided by the GDR ASR of CNRS and maintained by AlGorille. The RGE action organizes three meetings per year, and usually gathers 40-45 people per meeting.

Sylvain Contassot-Vivier decided to stop his animation role in the Embedded Pole in 2011 in order to focus on his research activities. However, he continues his expert analysis for the MEI (Mission d'Expertises Internationales).

ANR USS-SimGrid (2009–2011) and ANR SONGS (2011–2015)

Martin Quinson is the principal investigator of one project of the ARPEGE call from the ANR (french funding agency), called USS-SimGrid (Ultra Scalable Simulation with SimGrid – 2009–2011). It aims at improving the scalability of the SimG rid simulator to allow its use in Peer-to-Peer research in addition of Grid Computing research. The challenges to tackle include models being more scalable at the eventual price of slightly reduced accuracy, automatic instantiation of these models, tools to conduct experiments campaigns, as well as a partial parallelization of the simulator tool.

Martin Quinson is also the principal investigator of a project of the INFRA call from the ANR, called SONGS (Simulation Of Next Generation Systems – 2012-2016). It aims at increasing the target community of SimG rid to two new research domains, namely Clouds (restricted to the Infrastructure as a Service context) and High Performance Computing. We will develop new models and interfaces to enable the use of SimG rid for generic and specialized researches in these domains.

As project leading team, we are involved in most parts of these projects, which allows the improvement of our tool even further and set it as the reference in its domain (see Sections 6.2.1 through 6.2.5 ).

Bilateral Collaborations

With Arnaud Giersch from the University of Franche-Comté, we work on the design and implementation of a decentralized load-balancing algorithm, based on asynchronous diffusion, that works with dynamical networks. In such a context, we consider that the nodes are always available but the links between them may be intermittent. According to the load-balancing task, this is a rather difficult context of use.

Lucas Nussbaum and Martin Quinson are participating to a research effort lead by F. Suter from the Computing Center of IN2P3. This project is jointly funded by CNRS' Institut des Grilles and INRIA 's ADT Aladdin in a program that aims at bridging the production grids and distributed systems research communities. The overall goal of the project is to explore ways to enable the scientific study and the evaluation of improvements in the context of the gLite grid middleware. New results in this project are described in Section 6.2.7 .