Part of the INRIA ARC GREEN-NET Project : Power aware software frameworks for high performance data transport and computing in large scale distributed systems
The GREEN-NET is a Cooperative Research Action (ARC : Action de Recherche Cooperative) supported by INRIA. This project explores the design of energy-aware software frameworks dedicated to large scale distributed systems. These frameworks will collect energy usage information and provide them to resources managers and schedulers. Large scale experimental validations on Grid5000 and DSLLAB platforms will be proposed.
The Supercomputing 2008 Demo : "Turn your HPC green !" : The Good, The Bad and The Greener
Scenario 1 : the Good : an energy aware large scale distributed system user
||Bob an energy aware user, schedules a resource reservation on an 0 energy usage, ready to run platform. Through some energy sensors and visualisation frameworks, Bob can see the cost in energy of the resources he used. During his application execution, Bob monitors the energy used by his application in order to improve the patterns of his application.
See scenario 1 in lived snapshot (5 MB file)
Scenario 2 : the Bad : an energy un aware user
|| Lee, a Bad user, want to run its application at the time he requested. He schedules his reservations whatever the energy cost or no energy optimization that can happen.
Of course, Lee could have a look on the energy usage of the platform, but as he is a bad user, he doesn't care. ;-)
See scenario 2 in lived snapshot (7MB file)
Note : Bad users are not all bad, sometimes a BAD user is an urgent user that needs the resources for emergency and urgent applications . In that case the enrgy optimization in not to be taken inot account.
Scenario 3 : the Greener : an energy aware and energy enforcer user
Put your photo here
Tuco, a user aware of energy cost, wants to schedule a reservation. He is ready to read the Green advices proposed by the EARI systems and to follow the recommendations in order to reduce the global energy usage of the large scale distributed system. In the following scenario, he accepts to benefit of an aggregation with already scheduled reservations in order to reduce the energy consumed by the total reservations.
See scenario 3 in lived snapshot (12 MB file)
GREEN-NET Project Members
Laurent Lefèvre, INRIA RESO, LIP, ENS-Lyon, France
Anne-Cécile Orgerie, INRIA RESO, LIP, ENS-Lyon, France
Jean-Patrick Gelas, INRIA RESO, LIP, ENS-Lyon, France
Olivier Richard, INRIA MESCAL, Grenoble, France
Yannis Georgiou, INRIA MESCAL, Grenoble, France
- Jean-Marc Pierson, IRIT, Toulouse, France
- Georges Da Costa, IRIT, Toulouse, France
- Wu Chun Feng, Virginia Tech, USA
- Song Huang, Virginia Tech, USA
Inquiries may be directed to : Laurent Lefèvre (GREEN-NET project leader) - INRIA, Lyon, France - E-Mail: firstname.lastname@example.org