Call 2013 EoI

14-01-2014

Universita di Modena e Reggio Emilia

Research areas
  • Distributed computing
  • Situation awareness
  • Self organization & adaptation

www.agentgroup.unimore.it

Contact
Franco Zambonelli
Full Professor
Università di Modena e Reggio Emilia / Dipartimento di Scienze e Metodi dell'Ingegneria
Italy
Partner looking for project
Adaptive Machines in Complex Environments

05-01-2014

Babes-Bolyai University Cluj-Napoca, Cluj, Romania

Research areas

At the Department of Mathematics and Informatics there is a strong group of employees expert in different areas of Machine Learning, Artificial Intelligence, Adaptive Computing, Agent Modelling, Genetic Algorithms.

Research areas:

  • machine learning,
  • artificial intelligence,
  • genetic algorithms,
  • reinforcement learning,
  • data modelling for information extraction.

 

Web links:

  1. www.cs.ubbcluj.ro
  2. datamin.ubbcluj.ro
Contact
Lehel Csato
Associate Professor (Reader)
Babes-Bolyai University, Department of Mathematics and Informatics
Romania
Partner looking for project
Adaptive Machines in Complex Environments

03-01-2014

Transport and Telecommunication Institute

Research areas
  •  Transport and Logistics
  •  ITS  and Telecommunication technologies
  •  Applied Statistics
  •  Industrial Robotics and Robotics
  •  Simulation and Modelling
  •  Applied Statistical Methods
  •  Expert systems
Comment

Transport and Telecommunication Institute (TTI) is private high school located in Riga, Latvia. TTI is modern institution with old traditions in engineering area. Computer Science and Telecommunication faculty do research in area of electronics, industrial robotics, telecommunication, software development, etc. Telecommunication, electronics and robotics center of TTI equipped by most modern hardware and software solutions. Also we have positive experience in different EU level research programmes like FP, INTERREG, COST etc. To find more information: http://www.tsi.lv

We are looking for the project, which could be supported by our resources.

Contact
Mihails Savrasovs
Head of Research Department
Transport and Telecommunication Institute
Latvia
Partner looking for project
Adaptive Machines in Complex Environments

03-01-2014

Heterogeneous distributed system based on collaboration of energy-efficient processors and programmable logic

Keywords
FPGA, small, energy efficient processors, many-core processors, High Level Languages/Electronic System Level description, dynamic partial reconfiguration, big data processing
Abstract

Increasing resources offered by programmable devices have enabled their efficient utilization for large-scale data processing. Moreover System-On-Chip integration of the embedded processing system and the on-chip programmable logic creates unlimited possibilities to create custom accelerators that extend system performance. Furthermore, dynamic. partial reconfiguration (i.e ability to replace small fractions of the configuration without stopping the whole system) can be very powerful in many big data processing applications.

The main purpose of proposed project will be to build heterogeneous distributed system based on collaboration of energy-efficient processors and programmable logic. Using new technologies we plan to build a cluster of energy efficient nodes. We are going to propose new programming models employing high level languages integration with cluster data processing programming methods (like Map-Reduce). Ready to go hardware co-processors will be developed for data processing. By profiling the application, critical spots will be identified and special kernels will be implemented in dynamic reconfiguration grids. Furthermore, we would like to suggest techniques for optimization of streaming data across platform nodes, thanks lose-less data compression algorithms.

The CPU-FPGA couple will be supplemented with many-core accelerator. Many-core processing is an efficient way to process data in some application domains. 

The project aims to develop solution focused on energy-efficiency. The first benchmarks for system will be virus detection and plagiarism checker applications.

Partners already involved

ACC CYFRONET, Poland

Expertise needed / Role in project

We are looking for partners interested in the topic. Partners which are willing to collaborate in the ideas, software and IPCores development. We are experts in Xilinx's tool-chain but we don't limit ourselves to that technology.

Preferred countries
Different from those already involved
Comment

ACC Cyfronet computing offer is extended beyond traditional computing solutions. There will be utilized resources already existing (SGI RASC, Picocomputing platforms), as well, as clusters build directly to meet the needs of required tasks.

Contact
Sebastian Koryciak
Researcher
ACC Cyfronet AGH
Poland
Project looking for partner(s)
Heterogeneous Distributed Computing

02-01-2014

Green Multiscale Computing

Keywords
green computing, multiscale simulations, high performance computing
Abstract

Motivation: Investigation on efficiency of multiscale modelling dedicated to heterogeneous hardware architectures is currently performed in 2011/01/D/ST6/02023 national scientific project. Obtained results proved that for various multiscale calculations configuration of different control parameters, size of input data, load balancing and other algorithmic approaches result in different electric power consumption. Moreover, these calculations are finished within different periods of time, resulting in different consumption of electric energy, which highly influences costs of maintenance of advanced computer clusters and supercomputers.

Research hypothesis: It is possible to create computational strategy for multiscale numerical simulations, which, through optimal parameterization and scheduling of computational procedures, will allow to minimize costs of maintenance of computer clusters by reduction of electrical energy consumption.

Objective of research: The main objective of the project is creation of computational strategy allowing minimization of electrical energy consumption during performance of sophisticated multiscale numerical procedures. The strategy will take into account various configuration of input parameters and load balancing for single- as well as multiscale procedures. Such investigation will guarantee full spectrum of analysis of electrical energy consumption influenced by number of used devices, streaming processors, cores, computing time, etc. The strategy will be created for macro scale (Finite Element Method), micro scale (Cellular Automata), nano scale (Molecular Static and Molecular Dynamic) and for joined numerical procedures in different scales e.g. CAFE (Cellular Automata Finite Element) or FE2 (Finite Element Square)

Effect of the project: The new functional module of GridSpace2 platform dedicated to scheduling, performance and monitoring of virtual experiments by using grid infrastructure. The new module will implement energy-efficient strategy for performance of multiscale numerical simulations.

Methodology: Application of optimization procedures is crucial to develop the best energy-efficient strategy for performance of multiscale numerical simulations. In this project two types of optimization methods will be implemented i.e. single- and multi-objective. In the first case, methods already developed by Authors will be applied, especially parallel procedures based on nature-inspired approaches like genetic and particle swarm algorithms. In the case of multi-objective optimization, evolutionary methods will be analysed e.g. VEGA (Vector Evaluated Genetic Algorithm), HLGA (Hajela's and Lin's Weighting-based Genetic Algorithm), FFGA (Fonseca's and Fleming's Multiobjective Genetic Algorithm). The best of them will be selected, parallelized and adjusted to be used in grid infrastructure. The objective functions will be calculated by using single- and multiscale numerical procedures, while parameters of these methods and applied load balancing will be treated as optimization variables. Influence of various parameters on obtained results will be investigated by application of sensitivity analysis procedures e.g. Sobol, Morris and McKay methods. Due to applied optimization procedures and sensitivity analysis, it will be possible to determine strategy for multiscale simulations to obtain minimal time and electrical energy consumption simultaneously. Finally developed strategy will be tested on grid infrastructure for various multiscale applications, including procedures already used in GridSpace2 platform.

Influence on civilization: The research planned in this project is convergent with trend of green computing, which currently remains one of the most important challenges in HPC. More than twenty supercomputers listed on top500 requires electrical power of average wind farm (≈2MW). On the other hand similar energy consumption is required by the best six supercomputers in Poland. Therefore, each step leading to reduction of electrical energy consumption influences improvement of life quality.

Influence on science: Creation of the strategy will allow to constitute new basic rules for parallel and distributed programming, which will introduce energy-efficiency factor into scheduling, planning and load balancing procedures. Due to the new strategy, parallelization and distribution of multiscale simulations in heterogeneous hardware architectures will have to be reconsidered, especially for modern HPC environments.

Partners already involved

AGH University of Science and Technology - Poland

Expertise needed / Role in project
  • Researches involved in multiscale numerical simulations and high performance computing are needed
  • Anyone who is interested in investigation of parameterization of computational procedures on power consumption
Preferred countries
Different from those already involved
Comment

The description of workpackages is already prepared. If anyone is interested in extension of this document with new tasks and workpackages, I can send the description in doc format.

Contact
Lukasz Rauch
PhD
AGH University of Science and Technology / Department of Metals Engineering and Industrial Computer Science
Poland
Project looking for partner(s)
Heterogeneous Distributed Computing

02-01-2014

PGF Business Services

Research areas

<p>>PGF Business Services is a Startup founded in January 2013 by a group of Researchers in Engineering and Applied Mathematics, therefore we focus on innovative use of Mathematics and Numerical Methods within ICT and Green Energy Sciences.</p><p>We want to join R&D projects managed at various geographical levels, ideally EU funded projects, and want to be more and more involved in Research initiatives, always implementing Green strategies and sustainable approaches: our web site (<a style="font-size: 12px;" href="http://www.pgfbs.eu/">www.pgfbs.eu</a>) fully reflects PGF’s&nbsp; mission and scope.</p>

Contact
Pierluigi Fersini
R&D Manager
PGFBS
Italy
Partner looking for project
Heterogeneous Distributed Computing

24-12-2013

Kaunas University of Technology

Research areas
  • Computer science; ICT;
  • Software testing;
  • Smart systems.
Contact
Robertas Damasevicius
Vice-dean
Kaunas University of Technology / Faculty of Informatics
Lithuania
Partner looking for project
Heterogeneous Distributed Computing

16-12-2013

Polish Academic Computer Centre "Cyfronet"

Research areas
  • We are the computer centre that delivers computing capabilities to Polish science comunity
  • We are a competence centre for the reconfigurable computing. We deliver reconfigurable computing nodes to wider research community (SGI RASC and Picocomputing platforms)
  • We focus on energy efficiency in big data processing
  • We develope IPCores for floating-point high performance computing (e.g. exp(), sparse matrix)
  • We employ ESL tools and we are a official partner of Impulse-C Accelerated Technology
  • One of our research interests is run-time reconfiguration technic
  • Ar the moment we consider ZYNQ based cluster of heterogenous processing nodes
  • Our expertise is Xilinx's tool chain
  • We run the cluster of over 180 Tesla GPGPU cards in our computing centre
  • We develope algorithms for GPGPU (data processing, quantum chemistry)

http://www.cyfronet.krakow.pl/en/4421,main.html

http://www.cyfronet.krakow.pl/computers/13698,artykul,reconfigurable_platforms.html

http://www.cyfronet.krakow.pl/computers/13725,artykul,zeus.html

 

Comment

Imupulse-C, Xilinx FPGA, ZYNQ, big data processing, partial-reconfiguration, GPGPU, nVidia CUDA

Contact
Pawel Russek
Researcher
ACC Cyfronet AGH
Poland
Partner looking for project
Heterogeneous Distributed Computing

Pages