Project Title: Efficient Execution of Multidisciplinary Applications: New Challenges in the Multi/Many core Era.

Nacional

Oficial Identifier: TIN2011-28689-C02-01

Funding Entity: Ministerio de Ciencia y Educación

Main Researcher
     Tomàs Manuel Margalef Burrull

Researchers
     Ana Cortés Fité
     Carlos Brun Sorribas
     Julio Cesar García Vizcaíno
     Antonio Espinosa Morales
     Porfidio Hernández Budé
     Miquel Angel Senar Rosell
     Aprigio Augusto Lopes Bezerra
     Andrés Cencerrado Barraqué
     Joan Sorribes Gomis
     Anna Morajko
     Ana Ripoll Aracil
     Juan Carlos Moure López
     Eduardo César Galobardes
     Andreu Moreno Vendrell
     César Allande Álvarez
     Tharso De Souza Ferreira
     Claudio Daniel Márquez Pérez
     Claudia Andreina Rosas Mendoza
     Andrea Maria Martínez Trujillo
     Alvaro Chalar Zuñiga
     Arindam Choudhury
     Gonzalo Vera Rodríguez
     Abel Castellanos Carrazana
     Josep Jorba Esteve
     Gemma Sanjuan
     Tomàs Artés
     Alejandro Chacón

Research Support Technician
     Javier Navarro

Organizers
     Universitat Autònoma de Barcelona
     Universitat de Lleida

Project Structure
     The main goal of this project is the research on ideas, techniques, policies, and mechanisms for the efficient execution of multidisciplinary applications with high computing and data requirements on systems in the multi/many core era, including clusters, peta/exascale supercomputers, peer-to-peer systems, and cloud computing systems. The project is organized in four related layers:

  • Applications: Several grand challenge applications will be studied. Executing these applications efficiently on upcoming systems not only represents a great challenge from the computational point of view, but also for the development and evolution of the corresponding science or engineering research field (bioinformatics, environment, industrial, entertainment …).

  • Programming models and execution environments: They must be reconsidered both regarding these new architectural trends (computation, communication and memory organization) and the requirements of the selected applications. An analysis of the suitability of existing solutions and the development of specific extensions is mandatory.

  • Performance modelling, evaluation and tuning: Performance models are critical for efficient programming and tuning. They must be broaden or even redesigned to include the new features of current and future high-performance systems. Moreover, current monitoring and tuning mechanisms must be rethought to be effective in large-scale, multi/many core systems.

  • Resource Management Mechanisms: Distribution, assignment, and scheduling policies will be analyzed in connection with the different types of system and application considered. New proposals will be investigated for cluster, cloud and peer-to-peer systems, which also care about core-level management.