Plug in the grid and have all the computational power you likeParis 07 March 2002 Since several decennia, informatics have fundamentally changed the organisation of our society. Thierry Priol, Director of Research at INRIA, described in his talk at the DataGrid Conference in Paris how computational science, allthough fairly young and in constant interaction with mathematical and physical sciences, has been the object of numerous evolutions. The last development has been excessively focused in the media and involves a brandnew computing infrastructure, called the Grid in analogy with the infrastructures for distributed electrical energy.
To grasp the magnitude of the challenge, Mr. Priol tried to pin down the grid concept in a number of downright questions. Will researchers be able to apply computational power in the same way as we use electric power from day to day in our automated household devices? Which means without actually knowing where it comes from? This is one of the questions to which researchers today are searching the answer.
No need anymore to have a powerful PC at your home. The use of the Internet will enable us to access the necessary power made available somewhere across the network for calculations or storage without us really knowing where this power comes from, according to the speaker.
Today, this new challenge involves unexpected perspectives for the informatics research. In former days, most of the work concentrated on the management of a computer allowing us to use it efficiently, in the most secure and simple way. Some fifteen years ago, researchers started to take an interest in methods to program several computers simultaneously to speed up applications. This is generally known as parallel computing. From now on, with the apparition of the computing grids, a new dimension has occurred in the setting up of applications. We call it the network, as Mr. Priol stated. So the Grid is not completely new: it combines a number of already existing technologies.
The deployment of high-speed networks allows to conceive applications which are able to exploit several resources distributed geographically. The initiatives cited by Mr. Priol were Renater and VTHD in France and Géant in Europe. They give access to a variety of resources including supercomputers, databases, virtual reality systems or, quite simple, a great number of PCs.
The grid concept is slowly being materialised through mediatised applications like for instance SETI@HOME or, more recent still, Decrypton. In those two examples, the grid constitutes the free availability of PCs to perform specific calculations which are the search for extraterrestrial signals in SETI@HOME or the decipherment of the genome in Decrypton, as Mr. Priol explained.
Several tens of Teraflops have thus been acquired in a simple way by connecting PCs to the Internet. Ten years ago, this still represented an unimaginable power and caused people starting to dream, according to the speaker.
However, to obtain this power, extensive research is needed to make the grid transparent. All the traditional domains in informatics are involved including networks, exploitations systems, data management at high levels, security architectures, programming and execution models, computer infrastructures, algorithms and applications.
What exactly is the Grid? How can one classify the different modes of deployment? Mr. Priol presented the following classification scheme:
Priol sees two categories of Grand Challenges for the Grid. First is programming for the Grid. New algorithms have to be developed and code-coupling is an issue. Second category is the middle ware. The metacomputing and Internet computing paradigms are disctinct models that must be brought together. Perhaps we should work towards a Grid-aware operating system. The French ACI-Grid is tackling some of the Grid Grand Challenges. Computational Grids:
Grid programming:
Ad Emmen |
||||||||||||
 |