Computer architecture

Cloud Computing: A PEPR to reposition France in a strategic market

Changed on 10/04/2024
On 4 April 2024, the PEPR Cloud programme, co-piloted by the CEA and Inria, will be officially launched. The programme, which has a budget of €51 million over 7 years, focuses on several key issues: sovereignty, security and frugality.
PEPR Cloud
© Inria / Photo C. Morel

Supporting sovereign cloud research

From its emergence in the early 2000s to its mass adoption for storing volumes of data today, Cloud Computing is currently facing some major technological challenges.

Chief among these is Edge Computing, which involves maintaining a continuum between Clouds (hosted in data centres) and IoT devices through a massively geo-distributed federation of small data centres placed at the edge of the network.  A continuum of this kind is necessary to enable data to be processed as closely as possible to the way it is used, and to encourage the development of tomorrow's IT applications (smart cities, e-medicine, autonomous vehicles, etc.). It also offers a major opportunity to take back control of a market dominated by the GAFAMs, jeopardising Europe's sovereignty.

The software stacks developed for current Cloud infrastructures do not allow these changes to be taken into account at virtually all layers. Indeed, these solutions were not designed to incorporate the intrinsic characteristics of this new type of infrastructure, such as heterogeneity, the mobility of storage and processing resources, or the fact that these infrastructures are potentially managed by a number of different operators. 

"We find ourselves in a situation where these data centres, which were built on a large scale, no longer make sense in terms of our sovereignty and security issues. We are now faced with real problems of data management and responsiveness. Paris, with its thousands of cameras, is a case in point. If we need to process information from these cameras and analyse it quickly, we don't know how to do it, because it's not technically feasible to send all the video streams to a centralised data centre all the time", explains Adrien Lebre, director of the PEPR Cloud for Inria.

In addition, cross-functional aspects such as security and energy also need to be strengthened. From a security point of view, traditional mechanisms need to be adapted to take account of the increased attack surface induced by the Cloud-Edge-IoT continuum (more data, more heterogeneous data centres, more connected devices, mobility of users and resources, heterogeneity of access networks, etc.). 


We can no longer see these infrastructures as being built on a continuum. We have built something that is more and more beyond us every day.


Frédéric Desprez


Director of the Inria research centre at Grenoble Alpes University

As far as energy is concerned, Edge infrastructures offer more opportunities and leverage for action (renewable energies, reuse of energy dissipated by cooling, reduction in data movements, etc.) than large centralised infrastructures.  In addition, a number of issues will be addressed around taking into account and optimising the energy footprint of each application.

A PEPR to support changing needs in the field

This is a view shared by a number of players in the sector, including Inria and the CEA, which in 2021 have been entrusted with the task of steering the PEPR Cloud, launched by the government as part of the national Cloud strategy. The aim of this programme, which has a substantial budget of €51 million, is to continue and intensify the French Cloud research effort, as well as facilitating the transfer of innovations and solutions resulting from research to industry and the main open-source communities.


Technologies evolve, as do needs and demands... It is therefore important for France to accelerate research into Cloud Computing in the broadest sense, and in particular into the applications that we are building, or rather will build, on top of it (intelligent cities, industry of the future, tele-medicine, augmented reality, etc.). 


Adrien Lebre


Director of PEPR Cloud for Inria

The CEA and Inria, who decided when the PEPR was set up to work with representatives from the CNRS, ITM, UDICE and the CDEFI to draw up the programme's roadmap, have identified major areas of work: management of the lifecycle of infrastructures and the applications that run on them, supervision, data management across the continuum, new hardware and virtualisation, as well as two cross-cutting issues: security and controlling energy consumption.

"The aim behind these areas of work is to study all the software layers of the Clouds of tomorrow, from the lowest to the highest layers, and to do all this in a secure and frugal way: infrastructure supervision, data management, virtualisation management, information retrieval, energy management at all levels, etc. We have to take everything into account", explains Frédéric Desprez. 

A three-part roadmap

The first stage of the programme was to map the research community, which made it possible to draw up a complete inventory of the players involved in research dedicated to the Cloud, in order to link them to the subjects defined as strategic.

At the same time, seven priority projects were identified for this PEPR. Five of these concern the software layers needed to deploy the distributed Clouds of tomorrow, and two focus on the cross-functional aspects of infrastructure security and energy management, as mentioned above.

The PEPR roadmap also includes a call for projects focusing on two areas, the first on algorithmic work to optimise certain performance and reliability criteria for the software stacks developed, and the second on the impact of different application domains on the actual design of these software stacks. Finally, the programme includes a test bed in the form of a distributed experimental infrastructure. "The overall idea is to provide the sector with the keys to make a major transition, moving from a world where we have a few mega data centres controlled by the GAFAMs, to a more decentralised model with processing and storage resources as close as possible to usage," concludes Adrien Lebre.