Gabriel Antoniu

Senior Research Scientist, INRIA

Scientific leader of the KerData research team at INRIA Rennes - Bretagne Atlantique Research Center and IRISA.

Relevant research topics: parallel and distributed computing, distributed data storage and management, cloud computing, grid computing, P2P systems, transparent data sharing, fault tolerance in distributed data systems, data consistency, scalable deployment approaches for large-scale architectures.

My main current research interests are related to data management for large scale distributed infrastructures: clouds, petascale machines, desktop grids, hybrid architectures. I share these interests with Luc Bougé, with whom I co-founded the KerData Team.

Relevant topics are:

• Decentralized management of massive data on highly distributed infrastructures;
• Cloud data services;
• MapReduce-based data processing;
• Scalable transparent data storage and sharing;
• BLOB-based data management;
• Scalable distributed file systems.

A comprehensive sorted list of my publications and the corresponding PDF files can be found on HAL Open Archives Library: sorted by year sorted by type.

• New: latest interview about BlobSeer (Emergences, May 2012)
• General Co-Chair of the ScienceCloud 2012 International workshop held in conjunction with the ACM HPDC 2012 conference.
• Program Chair of the IEEE Cluster 2014 conference.
• 2nd Gilles Kahn PhD Award obtained in 2011 by Bogdan Nicolae, co-advised by Gabriel Antoniu and Luc Bougé.
• Editor for a Special Issue of Concurrency and Computation: Practice and Experience Journal on Cloud Computing for Data-driven Science and Engineering, 2012.
• Local Chair of the 7th International Workshop of the Joint INRIA-UIUC-ANL Lab for Petascale Computing, Rennes, June 2012.
• Track chair at IEEE CloudCom 2012 and IEEE CloudCom 2011 international conference.
• Local chair for the Parallel and Distributed Data Management topic of the Euro-Par 2011 international conference.
• Vice-chair of the NAS 2010 international conference for the storage track.
• Keynote speaker at MapReduce 2011 workshop (held in conjunction with ACM HPDC 2011), at the HPCS 2010 conference and at ADiS 2010 workshop (held in conjunction with CISIS 2010).
• Recent results in projects that I am leading as a PI:
  • Damaris shows excellent scalability up to 10,000 cores on the Kraken supercomputer for the CM1 tornado simulation!
  • A-Brain (AzureBrain) (2010-2012, project description) (PI: Gabriel Antoniu, co-PI: Bertrand Thirion, Parietal Project-Team, Inria Saclay – Île-de-France ), in collaboration with Microsoft Research within the framework the Microsoft Research - INRIA Joint Research Center. It is dedicated to joint neuroimaging and genetics analysis on Microsoft's Azure cloud computing platform.
    • Recent result (April 2012): The TomusBlobs BlobSeer-based storage-layer for Azure allows the target application to run on 350 Azure cores!
    • Live interview on TechDays TV: Cloud Story broadcast, Thierry Priol and Gabriel Antoniu, Microsoft TechDays, Palais des Congrès, Paris, February 2011.
  • MapReduce (2010-2014, project description). This project is devoted to using MapReduce programming paradigm on clouds and hybrid infrastructures. Partners: Argonne National Lab (USA), the University of Illinois at Urbana Champaign (USA), the UIUC-INRIA Joint Lab on Petascale Computing, IBM France, IBCP, MEDIT (SME) and the GRAAL INRIA project-team.
    • Result: BlobSeer enables faster concurrent VM multi-deployment (from 2 up to 25 faster than state-of-art). Results published at ACM HPDC 2011 (selection rate 12.9%)

• Initiator (with Luc Bougé) of the ENS-INRIA Prize for Excellency.
• PC member for the following recent international conferences (selection): ACM HPDC 2012 and 2013, ACM/IEEE SC 2013, IEEE/ACM CCGRID 2013, ICCCN 2012, IEEE HPCC 2012, IEEE AINA (2011, 2012) IEEE CloudCom (2010-2012), Euro-Par 2011, IEEE ICPADS (2010, 2012).
• Habilitation thesis published as a book:"L'approche JuxMem"

• Science Clouds Experiences: Sunny, Cloudy or Rainy? Panel Discussion in ScienceCloud '12: The 3rd Workshop on Scientific Cloud Computing, held in conjunction with ACM HPDC'12, Delft, June 2012.

• A-Brain: Using the Cloud for Neuroimaging and Genetics Research. Invited talk, Workshop on Trends in High Performance Distributed Computing, Amsterdam, March 2012. Held in conjunction with the PC meeting of the ACM HPDC 2012 conference.

• Advanced approaches for scalable data-intensive storage and processing. Invited talk, ORAP Forum on HPC, Paris, April 2012.

• Update on Damaris: How to Make CM1 Scale Linearly up to 10K Cores And What Comes Next. Invited talk, 6th International Workshop of the Joint UIUC-INRIA-ANL Lab for Petascale Computing, Urbana, USA, November 2011.

• Scalable cloud storage for data-intensive applications: the BlobSeer approach. Keynote talk at the 2nd International Workshop on MapReduce and its Applications (MAPREDUCE'11), San José, USA, June 2011. Held in conjunction with the ACM HPDC 2011 conference.

• BlobSeer: Efficient, Versioning-Based Storage for Massive Data under Heavy Access Concurrency on Clouds. Invited talk at Microsoft Research, Cambridge, UK, February 2011.

• BlobSeer: Enabling Efficient Lock-Free, Versioning-Based Storage for Massive Data under Heavy Access Concurrency. Invited Parallel@Illinois Seminar, NCSA, Urbana-Champaign, USA, April 2010.

• Autonomic Cloud Storage: Challenges at Stake. Keynote talk at the 2nd International Workshop on Autonomic Distributed Systems, Krakow, Poland, February 2010.

• BlobSeer is a data management platform we are currently developing for sharing massive data at very large scales. It originally relies on advanced techniques for decentralized data management and versioning techniques to provide scalable data throughput under heavy data access concurrency.

• Damaris: Damaris is a middleware for multicore SMP nodes allowing them to efficiently handle data transfers for storage and visualization by dedicating one or a few cores to the application I/O. It is developed within the framework of a collaboration with the Joint Laboratory for Petascale Computing (JLPC). It was successfully evaluated with the CM1 tornado simulation, one of the Blue Waters target applications, on several supercomputers (Kraken Jaguar), where it demonstrated excellent scalability.

• JuxMem: is a platform which illustrates the concept of Grid Data-Sharing Service, defined using a hybrid approach based on Distributed Shared Memory and Peer-to-Peer techniques.

• CoRDAGe is a generic co-deployment and re-deployment tool for dynamic grid applications.

• 2011 - Damaris has been officially validated for transfer to the Blue Waters project. Damaris has been preliminarily validated on up to 10,000 cores on the Kraken supercomputer (ranked 11 in Top 500) and up to 16,000 cores on Jaguar (ranked 3rd in Top 500).

• 2010 - BlobSeer:
  • On Azure clouds: BlobSeer has been integrated with Microsoft's Azure cloud platform within the TomusBlobs software service, in the framework of the A-Brain Microsoft Research - INRIA project (2010-2012, see below). Validated on 200 cores on the Azure cloud. Experiments on 1000 cores are in progress.
  • On Nimbus clouds: integration of BlobSeer with the Nimbus toolkit from Argonne National Lab in progress in the framework of the MapReduce ANR project (2010-2014).

• 2005 - JXTA: With Matthieu Jan, we proposed optimizations to the JXTA communication protocols, which substantially improved their performances (latency, throughput) on grids. Some were integrated in the official distribution of JXTA-C. This transfer is a consequence of performance studies performed in the framework of Mathieu Jan's PhD thesis (IEEE HPCC 2005, GP2PC 2005 - Best Presentation Award (held in conjunction with IEEE CCGRID 2005), IEEE IPDPS 2007).

• Seeding a France-Chicago Collaboration in Exascale Storage for Computational Science : (2012) FACCTS joint project with Argonne National Lab(ANL). Role: project co-PI, with Rob Ross (ANL).
• MapReduce(2010-2014, project description): an ANR project (ARPEGE 2010) with International partners on optimized MapReduce data processing on cloud platforms: Argonne National Lab (USA), UIUC, IBM France, the Joint UIUC-INRIA Lab for Petascale Computing, the GRAAL INRIA team, IBCP and MEDIT. Role: project leader. Project proposal
• F3PC: ANR-JST project (2010-2014). Role: local coordinator for INRIA Rennes - Bretagne Atlantique.
• Joint INRIA-UIUC Lab for Petascale Computing (JLPC). Role: leader of a joint research axis on concurrency-optimized I/O for post-petascale platforms.

• Projects with Tsukuba University, Japan (Osamu Tatebe, Gfarm team):
  • Bilateral PHC (ex-PAI) Sakura project (INRIA - AIST/University of Tsukuba, 2006-2007) on P2P-based data sharing. Role: project leader.
  • NEGST (2006 - 2009): CNRS-JST project. Role: participant.
• Bilateral project with the University of Illinois at Urbana Champaign, USA (CNRS-INRIA-UIUC programme, 2006-2007). Role: project leader.

• A-Brain (AzureBrain) project (2010-2012): INRIA - Microsoft Research project in the framework of the Joint INRIA - Microsoft Research Center. Role: project leader, with Bertrand Thirion (PARIETAL team, INRIA). It is dedicated to joint neuroimaging and genetics analysis on Microsoft's Azure cloud computing platform. More details: project description, press release.

AzureBrain in the news:

• Microsoft TechDays 2011 (Palais des Congrès, Paris, February 2011): Invited talk (Pierre-Louis Xech, Thierry Priol and Gabriel Antoniu) and live Web TV broadcast on MS Tech Days TV: ”Cloud Story” (8 February 2011 at 15:30)
• AzureBrain in "La Tribune"
• Cloud Magazine
• itespresso.fr

• IBM is our partner in the MapReduce ANR project (ARPEGE call, 2010-2014, see above). Topic: optimized MapReduce platform for clouds and hybrid infrastructures. Role: project leader.

• I led a Sun Microsystems – INRIA collaboration grant on Efficient use of Peer-to-Peer technologies on grid platforms (2005–2009). This grant funded Loic Cudennec's PhD thesis. Role: project leader.

• The SCALUS Marie Curie Initial Training Network, call FP7-PEOPLE-ITN-2008 (2009-2013). Role : coordinator for INRIA Rennes - Bretagne Atlantique (teams involved: KerData, Myriads). Other partners: Universidad Politécnica de Madrid, Barcelona Supercomputing Center, University of Paderborn, Ruprecht-Karls-Universität Heidelberg, Durham University, FORTH, Ecole des Mines de Nantes, XLAB, CERN, NEC, Microsoft Research, Fujitsu.

• CoreGRID ERCIM Working Group, since 2009. Role : main correspondent for INRIA Rennes - Bretagne Atlantique (teams involved: KerData, Myriads).

• DataCloud@work: an INRIA Associate Team with the University “Politehnica” of Bucharest (PUB), Romania (Valentin Cristea). Role: project leader.

• GridRand: bilateral PHC Brancusi project with the Technical University of Cluj-Napoca, Romania (2009-2010). Role: project leader.

• GridDataViz: bilateral project with “Politehnica” University of Bucharest (CNRS - Romanian Academy of Science, 2008-2009). Topic: visualization and remote control of the BlobSeer data management platform using the MonALISA monitoring framework. Role: project leader.

• CoreGRID European Network of Excellence (2004-2009). Role : participant.

• AE Hemera (2009 - 2014): INRIA Large Wingspan Action (Action d'Envergure). Role: leader of the Map-Reduce challenge. Co-leader of the Scalable Data Management Working Group.

• Grid'5000 and Aladdin-Grid'5000 Role: participant to the ACI GRID Grid'5000 project (2003 - 2008), then to the Aladdin-Grid'5000 project (since 2008). Our experiments use this experimental testbed on an everyday basis.

The Grid Data-Sharing Service approach I have worked on between 2004 and 2008 has been at the center of the GDS project of the French ACI MD (2003 - 2006) and has been enhanced and validated within the LEGO and RESPIRE ANR projects (2006-2009).

• GDS (2003 - 2006): ACI MD project. Role: project leader.

• RESPIRE (2006 - 2008): ANR project. Role: local coordinator for INRIA Rennes - Bretagne Atlantique.

• LEGO (2006 - 2009): ANR project. Role: local coordinator for INRIA Rennes - Bretagne Atlantique.

• GdX (2003 - 2006): ACI-MD project. Role: local coordinator for INRIA Rennes - Bretagne Atlantique.

• DataGraal (2002 - 2004): ACI GRID project. Role: participant.

A list of publications can be found on DBLP.

A comprehensive list of my publications can be found on HAL Open Archives Library: HAL (sorted by year) HAL (sorted by type)

• Viet-Trung Tran, Bogdan Nicolae, Gabriel Antoniu. “Towards Scalable Array-Oriented Active Storage: the Pyramid Approach”. ACM SIGOPS Operating Systems Review (OSR), Association of Computing Machinery (ACM), 2012, 46 (1), pp. 19-25.

• Bogdan Nicolae, Gabriel Antoniu, Luc Bougé, Diana Moise, Alexandra Carpen-Amarie. BlobSeer: Next Generation Data Management for Large Scale Infrastructures. Journal of Parallel and Distributed Computing, Volume 71, Issue 2, Pages 169-184, February 2011.

• Bogdan Nicolae, John Bresnahan, Kate Keahey and Gabriel Antoniu. Going Back and Forth: Efficient Multi-Deployment and Multi-Snapshotting on Clouds. The 20th International ACM Symposium on High-Performance Parallel and Distributed Computing (HPDC 2011), 2011, San José, CA, United States. Selection rate: 12,9%.

• Bogdan Nicolae, Diana Moise, Gabriel Antoniu, Luc Bougé, M. Dorier, “BlobSeer: Bringing High Throughput under Heavy Concurrency to Hadoop Map/Reduce Applications”, in : Proc. 24th IEEE International Parallel and Distributed Processing Symposium (IPDPS 2010), Atlanta, USA, April 2010. Electronic proceedings. Selection rate: 24%.

• Ramsés Morales, Sébastien Monnet, Indranil Gupta, and Gabriel Antoniu. Move:design and evaluation of a malleable overlay for group-based applications. IEEE Transactions on Network and Service Management, Special Issue on Self-Management, 4(2):107–116, 2007.

• Gabriel Antoniu, Loïc Cudennec, Mike Duigou, and Mathieu Jan. Performance scalability of the JXTA P2P framework. In Proc. 21st IEEE International Parallel & Distributed Processing Symposium (IPDPS 2007), Long Beach, CA, USA, March 2007. Electronic proceedings.

• Gabriel Antoniu, Hinde Bouziane, Mathieu Jan, Christian Pérez, and Thierry Priol. Combining data sharing with the master-worker paradigm in the common component architecture. Cluster Computing, 10(3):265 – 276, 2007.

• Gabriel Antoniu, Jean-François Deverge, and Sébastien Monnet. How to bring together fault tolerance and data consistency to enable grid data sharing. Concurrency and Computation: Practice and Experience, 18(13):1705–1723, November 2006.

Ph.D. students currently co-advised with Luc Bougé:

• Viet Trung Tran (2009 -)
• Matthieu Dorier (2011 -)
• Radu Tudoran (2011 -)

Ph.D. students currently co-advised with Maria Pérez (Universidad Politecnica de Madrid):

• Houssem Chihoub (2010-).
• Bunjamin Memishi (2010-).

• Alexandra Carpen-Amarie (2008 - 2011): PhD Thesis defended in December 2011. Now a Research Engineer at EIT ICT Labs.
• Diana Moise (2008 - 2011): PhD Thesis defended in December 2011. Now a Research Engineer at INRIA.
• Bogdan Nicolae (2007-2010): PhD Thesis defended in November 2010. Now post-doc at the Joint INRIA-UIUC Lab for Petascale Computing (JLPC), Urbana-Champaign.
• Loïc Cudennec (2005 - 2009): PhD Thesis defended in January 2009. Now Research Engineer at CEA LIST.
• Sébastien Monnet (2003 - 2006): PhD Thesis defended in November 2006. Now Assistant Professor (Maître de conférences) at LIP6, University Paris 6, in the REGAL INRIA Research Team
• Mathieu Jan (2003 - 2006) : PhD Thesis defended in November 2006. Now Research Engineer at CEA LIST

• Ecole Supérieure d'Informatique, Electronique, Automatique, 5th year, full grid and cloud computing module (since 2009), lectures (18h/year).
• University of Nantes, ALMA Master, Distributed Architectures module - AD (since 2009), lectures on grid, P2P and cloud data management (8-10h/year) and few hours of supervised work (TD) and practical sessions (TP) per year.
• ENS Cachan - Antenne de Bretagne, Master M2RI, PAP module (since 2004), lectures on P2P systems (10h/year).
• CEA-EDF-INRIA School (2009) on emerging grid middleware standards: lectures on grid data management (9h).
• INSA de Rennes, CS Department, 5th year, MPP module (2004 - 2009), lectures on P2P systems (2-4h/year) and 4h of practical sessions/year.
• University of Rennes 1, Operating Systems module (SYR), L3 level (2004 - 2008), lectures on networks (6h/year).

• Leader of the KerData research team since 1 July 2010.
• Scientific correspondent of INRIA Rennes - Bretagne Atlantique for international relations since (2008-2012).
• Elected member of INRIA's Evaluation Committee since September 2011.
• Member of INRIA's Committee for Scientific and Technology Orientation - International Relations Working Group (COST-GTRI) since June 2011.

• ACM HPDC 2012 international conference - Publicity Chair.
• HPCS 2012 international conference - Special Session & Workshop Chair.
• IEEE CloudCom 2012 international conference - Track Chair (MapReduce track).
• IEEE CloudCom 2011 international conference - Track Chair (MapReduce track).
• Euro-Par 2011 international conference - Local Chair for the Data Management Topic.
• IEEE NAS 2010 international conference - Vice-chair (Storage track).
• Euro-Par 2007 international conference - Publicity Chair.
• DRUIDE 2004 Spring School on Large-scale management of distributed data (national) - PC Chair

• PC member for the following international conferences: ACM HPDC (2012, 2013), ICCCN 2012, IEEE HPCC 2012, IEEE AINA (2011, 2012), IEEE CloudCom (2010-2012), IEEE ICPADS (2010, 2012), 3PGCIC (2010), Euro-Par (2008, 2011), IEEE Cluster (2004, 2008), IEEE CCGRID (2005, 2006, 2007, 2013), Euromicro PDP (2007, 2008), ICPP (2006).
• PC member for the following international workshops: MapReduce (in conjunction with ACM HPDC: 2010, 2011, 2012), ADiS (in conjunction with CISIS: 2010, 2011), DaMaP (in conjunction with EDBT/ICDT: 2009), HiperGrid 2009, HiPS (in conjunction with IEEE IPDPS: 2008), HPDGrid (2006, 2008), MSOP2P (2010), PDMST (2006).

With Luc Bougé and our American partner, Philip Hatcher (University of New Hampshire), I have initiated an associate research team abroad supported by INRIA, called Hyperion. The goal was to study the use of runtime systems for parallel and distributed computing on clusters and grids. This associated team ended in 2002.

My PhD thesis (available in french) proposed a generic implementation platform for multithreaded consistency protocols for Software Distributed Shared Memory: DSM-PM2.