D1 - Large scale systems

Person in charge
David BROMBERG (Professor Université de Rennes 1)
Description

Large-scale computer systems

LARGE SCALE SYSTEMS department addresses a variety of issues arising from large-scale computer systems featuring a large number of machines and/or users and/or data, typically in cloud, fog or IoT networks or in HPC.

Three main challenges

Research activities in this broad topic focus on three main challenges:

  • dependability and security;
  • performance, scalability and efficient resource management;
  • understanding, easing the use and programmability.

The main line of research

  • Models, resource and data management at scale systems: optimized data and metadata management in very large number of concurrent clients settings; reducing I/O overhead at very large-scales; scalability of k-nearest neighbor graph construction; improving gossip protocols with random linear network coding; modeling and reducing energy consumption in grids and HPC infrastructures
  • Convergence of HPC and big data: storage abstractions for workflows combining compute-intensive simulations and data-intensive analytics; modeling the behaviour of elastic storage systems for HPC infrastructures
  • System security: intrusion mitigation through privilege removal; machine learning approaches to intrusion detection; triggering unknown malware on Android; tracking flow of info in concurrent systems with a Coq proof; Models for quantitative analysis of security; Risk analysis based on attack trees
  • Cryptography: designing new and robust cyphers (block cyphers, lattice-based, homomorphic encryption, etc.); security proofs and security of cryptographic implementations (side channel attacks); formal proofs for cryptographic protocols.