Topic 7: Green High Performance Computing

Description

Optimizing power and energy consumption has been identified as one of the most critical issues on our way to exascale. Computations will have to be orders of magnitude more energy efficient than in today's architectures; applications will have to work with fixed total system power caps put in place to not exceed the limited power available; and systems will have to mitigate the impact of power swings during changing workloads. To achieve efficient execution of applications under these constraints, we require new approaches in all aspects of power-aware computing.

Recognizing the importance of these issues in high-performance computing, EuroPar is introducing for the first time a track on this topic, with the goal of providing a forum to bring together researchers in this wide field. Topics include techniques at all levels of the hardware/software ecosystem, including the architecture, the middleware, the programming model and the application, as well as approaches in power measurement, analysis, modeling and prediction.

Focus

• Power-aware scalable architectures
• Power-aware parallel programming models, languages and runtimes
• Power efficient algorithms and applications
• Performance optimization under a power bound
• Power measurements and measurement tools for systems and applications
• Power and energy modeling and prediction of systems and applications
• Power-aware job scheduling
• Power optimization techniques for large scale I/O
• Tradeoffs between power and resiliency

Topic Committee

Global Chair
Martin Schulz, Lawrence Livermoore Nat. Lab., USA

Local Chair
Luís Lopes, University of Porto, Portugal

Further Members
Enrique S. Quintana Orti, Universidad Jaime I, Castellon, Spain
Koji Inoue, Kyushu Institute of Technology, Japan