About project

Project title:Modeling efficiency, reliability and power consumption of multilevel parallel HPC systems using CPUs and GPUs
 
Principal investigator:Paweł Czarnul, PhD DSc Eng
 
Project number:2012/07/B/ST6/01516
 
Financing institution:National Science Center, Poland
 
Performing institution:Faculty of Electronics, Telecommunications and Informatics, Gdansk University of Technology
 
keywords:distributed large scale computational systems, parallel and distributed applications, scalability, reliability of computations, power consumption of large scale systems, CPU, GPU

Goals and effects

  1. Creation of parametrized models of large scale systems (volunteer and cluster based) which consist of components such as computational (CPU, GPU) and network devices.
  2. Definition of models of parallel applications that can ba run in the aforementioned systems.
  3. Definition of functions that estimate execution time, reliability and energy consumption of computational and communication blocks in an application based on practical tests. These functions are used by a simulator.
  4. Implementation of a simulation environment for simulation of execution of parallel applications on large scale systems. The environment consists of:
    1. client application – editor that allows modeling systems and applications as well as simulation management panel. It is possible to define a system model, application model (as a Java application with extensions representing computational and communication blocks)
    2. hardware database (includes computational and network devices)
    3. models that define execution times of blocks, energy consumption and reliability
    4. simulator of execution of the aforementioned parallel applications in the system – it allows conducting simulations that return application execution time, used energy and simulated reliability of execution.