Technical Platforms of Intensive Computing

The intensive computing or high performance computing (HPC : High Performance Computing) is a set of techniques and means ( computers, softwares and expertise) intended to deal with complex applications (simulation, design and scientific computation, ...) by using computers specializing in the fast processing of large volumes of digital data.

The fields of application of intensive computing are many, its mastery and its use have become indispensable. It has become an essential tool both for researchers and industrialists and practically in all disciplines. It can be found in :

• Hydrocarbons: tank modeling, seismography, discovery, and from refining to distribution.
• Industrial applications: fluid mechanics, assembly lines, reduction of vehicle weights, aircrafts, missiles, design and validation of cars,...
• Animations: film making (fires, water, surfaces ...).
• Creation of virtual world: games, medicine, military industry,...
• Daily life: prediction of customer behavior in supermarkets (example: correlation of a product A with another product B).
• General Relativity (solution of Einstein's equations).
• Resolution of equations (differential, partial derivatives ...).
• Molecular dynamics.
• Quantum chemistry.
• Astrophysics.
• Climate modeling.
• Weather.
• Finance.
• Etc, ...

Therefore, the nature and the profile of the means of calculation required by the scientific challenges are growing day by day, hence the need to increase the national calculation capacities by improving the power, the speed of processing and the storage

At present, research and development, deployment in international markets, industry, defense, etc., are measured by the capacity to intervene, in different fields, within optimum deadline, which means that each State tries to repositioning in terms of digital computing means. Algeria, aware of these challenges, confirms its ambition to resume its position and to be a regional leader in the medium term by strengthening these computing capacities.

This will be achieved through the deployment of intensive computing platforms at higher education and research institutions.

In the industrial environment the requirements are well defined and often unique, and in the laboratory environment, the requirements are multiple and heterogeneous by their natures which are related to research activities covering a multitude of disciplines: mathematics, physics, chemistry, medicine, engineering, different sciences...
In the first case (for industrialists) the choice and sizing of equipment and software are relatively simple, and the budgets allocated can be enormous given that there is a return on investment.
On the other hand, for laboratories where the diversity of scientific applications is specific to each and therefore important, the choice of a hardware and software solution is at least more complex, plus budgetary constraints.

In order to satisfy a majority of researchers in the immediate future, the Directorate-General for Scientific Research and Technological Development (DGRSDT) has allocated the necessary funds for the creation and equipping of twenty-five (25) Technical Platforms of Intensive Computing and decided to support the institutions concerned by this operation by bringing them together and by calling on specialists to define together a set of specifications that ensures, on the one hand, the different needs of each other and, on the other hand, the choice of a suitable architecture and to homogenize the technical characteristics of the equipment and software in order to ensure their compatibility for a possible use in a regional or national computer grid through the ARN network.

To this end, a working group was set up on the initiative of the DGRSDT, bringing together all the players involved in intensive computing in order to draw up a single set of specifications.

The researchers using Intensive Computing representing twelve (12) institutions have adopted a scalable common architecture, based on the latest technologies made up of several servers organized in clusters. This cluster consists of several dual-core and multi-core nodes, with 2GB of memory per core and 500GB of hard disk for each node, Giga Ethernet and QDR InfiniBand network interfaces and a 15TB storage capacity as well as accessories (cables, optical jumpers, KVM, screen,...) where the set is assembled in a 42U cabinet, 19". A cluster management and management platform based on free and standard software as well as the different libraries and compilers have been stopped.
The first beneficiaries of the acquisition are the universities of Oran, Houari Boumediene, Guelma, Constantine, Tlemcen, Sétif, Sidi Bel Abbes, Tizi-Ouzou, Annaba, Médéa, ENP of Oran and CERIST.

In a second phase, this list was extended to thirteen (13) other universities: Chlef, Skikda, Laghouat, Batna, Béjaïa, Biskra, Mostaganem, Mascara, Bordj Bou Arréridj, Ouargla, Djelfa, USTO and Jijel. The mapping of the distribution of these technical platforms of intensive computing is as follows :

(Mapping of the Technical Platforms of Intensive Computing)

Within the framework of the policy of valorization and connection between university-industry, the Ministry of Higher Education and Scientific Research and especially that of the DGRSDT, institutions endowed with different common services, must commit themselves to develop collaborations with the industrial and socioeconomic world and examine all possibilities of cooperation: transfer of knowledge, audit, and expertise.
A wide range of competences exist in laboratories and research centres, multidisciplinary teams must be set up to make available to researchers and trainees as well as industrialists these means of calculation by accompanying these users (confirmed or not) in different stages of their work.
These technical platforms of intensive computing will provide all related services: software acquisition assistance, installation and configuration, support users, informing and making available to users the existing solutions, developing specific calculation codes, storing data.