French ACI Globalisation of Computing and Data Resources project refines grid computing concept

At the same time, Mr. Cosnard continued, the Globus system was developed to connect American supercomputers and thus build an even more powerful parallel computer. In this case, we are not talking client-server technologies anymore since all computers can perform calculations and use the power of all the other ones. The applications show a far greater degree of complexity in performance. The European projects Datagrid and Eurogrid are using this type of technology.

These models and algorithms are the most well known and frequently used allthough they only constitute the first and very primitive versions of grid computing models, according to Mr. Cosnard. The ACI Globalisation of Computing and Data Resources (GRID), established in February 2001 by Roger-Gérard Schwartzenberg, the French Minister of Research, aims at the development of research activities in this domain to invent and implement new models. The GRID objective is to activate via the Internet the computing power and storage capacity of installed computers. This "super web" intends to respond to the computing and storage needs of scientists. The GRID action forms a real dynamic experimental tool pushing research to distributed programming environments, high-performance networks, and new natural language access interfaces.

Funded by the French National Science Foundation, the GRID action will increase the national research effort in the domain of Globalisation of Computing and Data Resources. The ACI objective consists in boosting and making operational the contribution of French research teams in this domain by supporting teams that are already active, by attracting new actors, and by stimulating contacts between developers of new solutions and the users, as anticipated by the ACI GRID manager.

Using high-performance network platforms provided by RENATER and the high-performance network VTHD, developed by RNRT, the national telecommunications research network, and relying on regional or academic infrastructures, several research initiatives will be developed to contribute to the implementation of "experimental grids", Mr. Cosnard told the audience. These initiatives consist in the deployment of algorithms for distributed access, authentification, security and exchange handling of resources, interfaces for "metacomputing" or the exploration of information layers, decision-making, etc.; and systems and environments for distributed computing or for the exploitation of large data, modelling, algorithms, code coupling, visualisation, and pre- and post-processing.

GRID aims to contribute to the creation of computing tools that allow the implementation of an "ambiant network" with a transparent access to "virtually unlimited" computing resources and to the foundation of enterprises, particularly in the domain of application service providers (ASP). Another objective, expressed by Mr. Cosnard, is allowing French research teams to position themselves at an international level, and in the first place a European level to prepare answers to the Sixth European Research Programme for Information Technologies.

The application scheduling on computing grids is difficult because the computing resources are distributed geographically and reside on parallel supercomputers. Globus enables to mask the distributed nature of these resources by considering the grid as a virtual parallel supercomputer whose programming is done via communication libraries based on exchanging messages. Unfortunately, it is not possible to easily use this process for all applications that could use the grid, as Mr. Cosnard explained. A possible approach which will be developed in the ACI GRID framework, is to combine the parallel computing models based on tasks and messages, and the distributed computing models based on distributed Java objects and Corba components. In this approach, a grid application constitutes an entity of objects or components in interaction and these objects could be parallel programmes themselves.

Global Grid computing projects such as SETI@home and Entropia essentially offer the computing capacities of desktop computers that are voluntarily used for applications of a trivial parallelism. On the other hand, Peer to Peer systems such as Napster and Gnutella use the storage resources of voluntary PCs. However, there exists no Peer to Peer system with an infrastructure allowing to launch calculations on participating PCs, Mr. Cosnard told the audience. This approach which will be developed in the ACI GRID, aims at increasing the functionalities of Global Grid systems by integrating the possibility to use storage resources of participating PCs and by offering the chance to no matter which participating PC to independently launch calculations on the system.

An approach of the highest level is based on the Application Service Provider (ASP) concept in which clients remotely access computing resources. Here, a series of servers is deployed on a computing grid providing access to clients distributed over the network. The aim is to enable remote access to high-performance machines with a huge memory capacity without the need for the user to be an expert in parallel computing, nor to install complex algorithms or hermetic interfaces on his machine, but being able to use in a secured manner confidential codes and data on the servers of those who develop them, as Mr. Cosnard stated.

Currently, the information systems contain large volumes of data in various formats. Data grids enable to distribute these data on a wide range of devices which dynamically evolve and are accessible to the greatest number of users. To maintain these grids, systems are needed which allow to access multiple sources of pre-existent autonomous and potentially heterogeneous data. Therefore, an infrastructure is required which offers transparent access to a wide range of resources, making the complexity of the underlying architecture transparent to the user. Such a system offers features to conceive, integrate, query, and manage data between the various applications in a save and efficient manner.

Data and computing grid applications can be found in all scientific areas. One of the ACI GRID goals therefore is to establish multi-disciplinary consortiums who should be able to solve new or complex problems, as Mr. Cosnard concluded.