INCITE was conceived by the Department of Energy and in particular the Office of Science. The acronym stands for Innovative and Novel Computational Impact on Theory and Experiment. To get the best insight into this programme, one should read Dr. Raymond Orbach's comments. The key driving thrust was to build machines of competitiveness.

There is a major shift in thinking and all areas of science and technology are in direct focus, explained Dr. Kalwani. Stating his point, he cited Dr. Orbach: "What we are seeing now is a transformation in the way we do computation ..." and then in conclusion: "It's a new sociology for science, for achievement, for discovery."

At a personal level the Council for Competitiveness was a major driver in assisting and making this viewpoint popular amongst the industrial partners, testified Dr. Kalwani. The Council for Competitiveness brought industry and academia together.

There are actually two major areas through which this is realised. Initiated in 2003, the programme provides the Office of Science computing resources for a specific set of "highly" computationally intensive research projects of a very large scale, that cam make high-impact scientific impact through the use of major type of equally large allocation of compute power.

The first major award was in 2003. Back then, it got started with 4,9 million CPU hours and was shared among 3 projects, the speaker told the audience. The term "processor-hours" refers to the allocation of time on a very large system:

• in 2003 4,9M CPU hours for 3 projects
• in 2008 265M CPU hours for many diverse projects in chemistry, biomedicine, etc.

Dr. Kalwani partnered with the company Fluent Inc. in a project called "CAE Simulation of Full Vehicle Wind Noise and other CFD Phenomena". In 2007 INCITE allocations were offered to Boeing, Dreamworks, Pratt and Whitney.

The speaker gave an overview of the INCITE test problems:

Test no. 1 involved a full vehicle wind buffeting simulation. The full vehicle model included exterior and interior surfaces. It was a transient, LES turbulence model showing 18-24 Hz Hydrodynamic Phenomena. The wallclock time on present day resources by way of comparison would take more than 4 weeks. The no. of cells involved several million ones.

Test no. 2 involved a semi-trailer truck pass by simulation. The numerical procedure simulated a 70 mph passing event in a transient, realizable K-c turbulence model.

Test no. 3 involved a fuel filling simulation of a complete fuel tank and filler tube assembly. The researchers utilized special custom home grown codes for fuel vapor creation, a transient model in nature. The wall clock time on present resources would take several weeks because it is a very hard problem. The fuel splashes around in the tank which is dangerous because of the different temperatures in the tank. The problem to solve is how this can be made more safe.

Test no. 4 was about a Buoyoncy Convection Simulation: the convection caused by temperature gradient in hot simplified underhood compartment of a stationary vehicle. This was tested with a transient K-c turbulence model in an ideal gas law.

Test no. 5 consisted of a fluid immersion and drainage simulation: the volume of fluid approach to interface tracking in a transient model with dynamic moving mesh to describe the kinematic motion. The current resources for this are unknown because it is an even harder problem. The immersion phase alone requires several weeks.

Test no. 6 constituted a paint oven drying simulation with conduction, convection and radiation heat transport in a transient model to simulate the dynamic moving mesh to describe the kinematic motion.

In all this the biggest challenge was, to Dr. Kalwani's great surprise, not technical but concerned the legal issues. Six months were spent on solving them.

Other items were also looked at, including fuel cells, entirely new synthetic materials, and complex combustion processes. Several systems were involved, delivered by NERSC, ANL, ORNL, etc.

Right now there are plans for exascale and beyond with a Call for Proposals in 2009. The projected INCITE hours are planned as follows:

• 2011 10 billion hours
• 2014 100 billion hours
• 2019 1000 billion hours

In the discussion afterwards with moderator Marc Parsons from EPCC, Dr. Kalwani, Dr. Hamelin and Dr. San Luis, it came about that capacity versus capability is a big issue in Europe. Drs. Hamelin and San Luis stated that they need both. Dr. Kalwani however urged to focus on capability, since the capacity will follow.

Asked for the INCITE usage percentage in industry and academy, Dr. Kalwani said that is was more than 5 percent, at present even 8 percent, and still growing. PRACE needs to do a better job in educating the government, he warned but confirmed that INCITE does work. It is altogether not a new phenomenon.

Dr. Kalwani philosophized that government is a reflection of society. Society must realize that this is the right thing to do. PRACE must find volunteers to take up the task and passionately believe in it.

In answer to a question from the public Dr. Kalwani said that having the data spread over geographically distributed centres has advantages to save mission-critical data in periods of disaster.

He also stated that the TOP 500 does not show how we get real science out of the supercomputers. It is only linpack. Solving 2 billion equations in 2 seconds is not relevant if we forget the real science you can get out of it.

Universities have now courses on computational science so students can learn to understand the problems of physics and maths and the issues industry is dealing with, Dr. Kalwani concluded.