Anne-Marie Kermarrec, researcher at INRIA Rennes-Bretagne Atlantique is awarded a 1.2 million euros l'ERC starting grant. Beyond the fact that this acknowledges the quality of her research, this grant provides her with a great practical and intellectual stability for the 5 next years. Anne-Marie Kermarrec has been selected with 297 other researchers, among 10,000 applications in Europe, all sciences included. This success is due to several factors: major contributions in the field of distributed computing, an European career (one year in Vrije Universiteit in Amsterdam and four years at Microsoft Research in Cambridge, UK) and an original, ambitious project clearly fitting within the "high-risk high-gain" spirit of the call.
Gossple aims at radically changing the navigation on the Internet by placing users affinities and preferences at the heart of the search process. Complementing traditional search engines, Gossple will turn search requests into live data to seek the information where it ultimately is: at the user. Anne-Marie Kermarrec favourite example to illustrate the limitations of the state of the art solution is the following: consider a family living in Rennes looking for an English speaking student who would be willing to trade baby-sitting hours against accommodation. Given the high number of students in Rennes, there is no doubt that such an offer would be of interest for many English speaking students. Yet, satisfying this simple, slightly unusual, request is challenging and in fact almost impossible using search engines. The data is clearly out there but it is difficult to achieve the match between the offer and the supply. One solution would be to contact by e-mail all the English speaking students of the city, should this information be available.
Gossple precisely aims at providing a fully decentralized system, auto-organizing, able to discover, capture and leverage the affinities between users and data.
Complementing Google-like search engines, Gossple will turn the request into a dynamic object navigating the network using epidemic protocols to find matching users/data. relevant clusters of users and data. At the heart of this procedure lies dynamic overlays based on users affinities, preferences and recommendations.
This goes far beyond discovering indexed data. Why is Gossple so challenging scientifically speaking? In a nutshell this can be summarized in four words: decentralised nature, scalability, adaptation to dynamics and reactivity. Providing distributed algorithms that let global properties emerge from individual decisions, connecting billions of users and/or objects and navigating efficiently in such extremely large digital universe, capturing and leveraging affinities, habits, recommendations, coping with arbitrary behaviours, providing efficient protocols to match the offer and the supply are as many difficult issues to be solved.
In short, the expected outcome of Gossple are theoretical advances in large-scale distributed computing, practical advances in large-scale distributed systems, an innovative software to create dynamic autonomous networked systems and the ground for new kind of distributed collaborative applications.
Epidemic protocols and peer to peer overlay networks are at the core of Gossple and the expertise and activities of Anne-Marie Kermarrec and more generally her collaborators from the research group ASAP (As Scalable As Possible) that she founded in 2007.
The ERC grant for such a project is a clear sign of the importance of the research addressed by this newly created research group ASAP from INRIA Rennes - Bretagne Atlantique.