OPAL, a shared platform for high performance computing

The OPAL (Observatoire Pluridisciplinaire des Alpes-Maritimes) project is part of the movement to structure the region's higher education and research players around the Université Côte d'Azur.

OPAL is based on the pooling and development of high-performance computing and visualization facilities at Inria Sophia Antipolis - Méditerranée, Université Côte d'Azur (Azzurra mesocenter), the Observatoire de la Côte d'Azur (SIGAMM mesocenter) and Mines ParisTech (Laffitte cluster).

In addition to funding from Inria, the Côte d'Azur Observatory and Mines ParisTech, OPAL is supported by the French government (DRRT), the Provence-Alpes-Côte d'Azur region and the Sophia Antipolis urban community (CASA) as part of the 2015-2020 CPER (State-Region Plan Contract). OPAL is also receiving funding from the IDEX UCAjedi for the Azzurra equipment.

Actions over time:

• participation in all the actions common to the project partners (2014-2020): governance, funding applications, working groups, common infrastructures, communication ;
• evolution (2017-2019) and integration into OPAL (2020) of the Nef computing cluster equipment, positioned as a scientific experimentation infrastructure for computing, massive data and visualization;
• complete renewal (2018-2020) and integration into OPAL (2020) of the 3D immersive visualization equipment Gouraud-Phong.

The new Gouraud-Phong 3D immersive visualization room is composed of :

• a 16:9 ratio screen, 3.05 * 1.77 meters in size;
• a touch screen ;
• a stereo laser projector, 2 560*1 440 pixels, 32 contact points, object tracking solution in a volume of about 4x2x2 m³ in front of the screen by 6 low latency infrared cameras;
• a high speed (40 Gb/s) and low latency (<1.5 μs to cross the 200 m of the Inria campus between the graphics rendering equipment and the Nef computing servers) Infiniband network connection allowing to supervise and interact on numerical simulations by a real time coupling between visualization and high performance computing.

The pooling of equipment within OPAL allows for a progressive thematic "coloring" of the equipment in order to better cover the specific needs of certain research communities at the Côte d'Azur academic cluster level.

In this context, Nef is the OPAL facility that is used extensively by the digital science communities, including parallel computing, digital simulation, artificial intelligence, deep learning and data sciences. Nef is primarily aimed at experimentation, code development and application prototyping. In technical terms, Nef's recent developments have focused on gas pedals (150+ GPU cards) and massive storage (600+ TiB BeeGFS). As a result, beyond Inria, Nef is particularly attractive for projects from the 3IA Côte d'Azur, or research teams from the I3S laboratory at UCA.

In 2020, the project partners finalised the implementation of cross-access to OPAL equipment: signing of a multi-partner framework agreement, drafting of a user charter, establishment of rules for managing sensitive or personal data, definition of a platform citation policy, creation of a technical repository of right holders, creation of the OPAL website. These actions led to the opening of OPAL on 30 September 2020.

To meet the most pressing needs, 65 users have been given access to Nef in advance of the official opening, demonstrating the interest and relevance of the OPAL approach. Moreover, the Nef equipment is open to all Inria collaborators. By promoting the sustainability and evolution of Nef, OPAL also contributes to the pooling of computing resources at the Inria level.

The OPAL project has enabled the creation of a digital experimentation tool for academic players in the Côte d'Azur region, encouraging transdisciplinary research within the Université Côte d'Azur and its partners. OPAL is the concrete expression of a proactive approach by the site's academic players and their long-term investment in a common structure. At the level of Inria, OPAL demonstrates the capacity of all the players to mobilise around a centre objective.