 Download the podcast on iTunes or listen now 
Modern scientific research requires a huge amount of computing power. As supercomputers are expensive, another option is to take advantage of the huge amount of unused computing power from desktop computers. Indeed, research institutions, laboratories or companies often have a huge amount of unused computing power, sometimes distributed over hundreds of desktop computers used by the staff. Distributed computing exploits the computing power of these desktops during idle time.
The ACT-DC project
Three years ago, triggered by the success of the SETI@home project, the Advanced Concepts Team (ACT) of the European Space Agency (ESA) started to work on the development of a distributed computing environment to be used within the ESA computer network: the Advanced Concepts Team Distributed Computing environment (ACT-DC).
The idea behind the ACT-DC is to install a programme, such as a screensaver, on the office computers of volunteers. When a computer is not being used and the screensaver is activated, this program communicates with the server via the Local Area Network. The server continuously breaks down computing tasks into subproblems, distributes them among the available clients and generates the final results by assembling the received solutions. A client computer only asks for subproblems when there is no user activity detected on the mouse or keyboard.
This is the so-called idle time computing concept, which has been successfully used in famous projects such as SETI@home.
The test phase
 The first official test of the ACT-DC was made on August 2005 at ESTEC, ESA’s research and technology establishment. The test lasted 9 days and involved 40 client computers. A 70 GFlops performance index was reached during the test. The test aimed at solving a problem related to ionosphere data processing.
ACT-DC application
Since this successful test, programmers and researchers at the Advanced Concepts Team have been using the ACT-DC mainly to solve an important computationally hard problems, such as global optimisation problems. Global optimisation problems arise daily in almost all operational and managerial phases of engineering processes and in particular of space missions. It is thus essential to find unbeatable solutions to optimisation problems of any kind. The ACT-DC has already relied on a number of highly distributable global optimisation solvers, such as Particle swarm optimisation, simulated annealing, differential evolution, genetic algorithm and Monte Carlo search.
This research will help in applications such as interplanetary trajectory optimization, the design of gridded ion engines, artificial joint design, analysis of the orbits of potentially hazardous near-Earth objects (asteroids), Earth observation data analysis, design of advanced micro-structured radiators, etc.
The future
In the near future, a large-scale demonstration of the distributed computing environment - 'the big run - will be held. The team is also moving towards making the distributed computing environment a completely decentralized architecture. Rather than relying on a single server, it will share and pass information between peers (similar to Bit-torrent file sharing is already doing). Developments on the DiGMO central intelligent server are also under way to improve its autonomous decision capabilities.
|
For ESA employees: if you would like to contribute to the distributed computing environment, you can install our beta version of the screensaver. To do this download this .exe file, place it on your desktop and double click on it.
People involved in the development/test of the ACT-DC and of DiGMO:
Working in the ACT:
Dario Izzo (Project coordinator), Dario.izzo@esa.int
Marco del Rey Zapatero (YGT in Artificial Intelligence)
Previously employed at the ACT:
Tamas Vinko (research fellow in Informatics)
Mihaly Csaba Markot (research fellow in Informatics)
Fabio Pinna (stagiaire)
Non-ACT:
Angela Aragon Angel (YGT for TEC-ETT)
Further information
Our publication pages provide some further reading
D. Izzo and M.Cs. Markót: "A Distributed Global Optimisation Environment for the European Space Agency Internal Network."
Proceedings of the GO 05 – 4th International Workshop on Global Optimization.
Izzo D., Markót M.Cs., Nann I.: “A distributed global optimiser applied to the design of a constellation performing radio-occultation measurements”,
Paper AAS 05-150, Proceedings of the 2005 AAS/AIAA Space Flight Mechanics Conference, Univelt Inc., Vol. 121, pp. 739-748, 2005.
Rahtke A., Izzo D.: “Options for a non-dedicated test of the Pioneer Anomaly”, Journal of Spacecraft and Rockets, Vol. N
|
Back
