Étienne MAUFFRET

Doctoral student, Polytech'Annecy-Chambery.

Contact

E-mail: Etienne.Mauffret- @ -univ-smb.fr

Telephone: +33(0) 4 50 09 42 09

Fax: +33(0) 4 50 09 65 59

Office : Axxx

Address: LISTIC - Polytech Annecy-Chambery, BP 80439, 74944 Annecy le Vieux Cedex, France

Etienne Mauffret has been a doctoral student at université Savoie Mont Blanc since October 2017. He teaches at Polytech' Annecy-Chambéry and carries out his research at LISTIC (where he is a member of the laboratory board).

GroupNetworks and Distributed Software Systems

Publications : referenced on HAL

Thesis

Topic: Large-scale distributed data management - placement, indexing, modular consistency and fault tolerance

Abstract: This thesis is part of the ANR RainbowFS project, whose consortium is composed of: (i) Université Pierre et Marie Curie (Paris VI), (ii) Université Savoie Mont Blanc, (iii) Laboratoire d'Informatique de Grenoble (LIG), (iv) Télécom SudParis (TSP), and (v) the industrial company Scality. The thesis will be carried out in the LISTIC laboratory at theuniversité Savoie Mont Blanc. It will be directed by Sébastien Monnet (LISTIC/USMB).

The aim of the RainbowFS project is to design and implement an approach to data management that delivers the right level of consistency, at cloud scale. Existing systems force applications to use predefined consistency models. The RainbowFS project offers tools for co-designing applications and their consistency protocols. The aim is to offer data availability while preserving consistency guarantees (respect for invariants). Many operations can be designed to run asynchronously, with synchronization taking place when strictly necessary.

This thesis will focus primarily on data placement, localization/indexing, consistency and fault tolerance. Indeed, all these issues are closely linked. Data is geo-distributed and replicated. The various copies of a data item must be placed (and moved) judiciously, taking into account access (frequency and type) but also criteria of fault tolerance (in order to cope with correlated faults, for example), confidentiality and so on. When copies are placed according to precise criteria, it is not possible to use an implicit localization mechanism such as can exist in distributed hash tables (DHT), so it becomes necessary to design indexing/localization mechanisms enabling efficient data access. Moreover, in large-scale systems, nodes arrive and leave frequently. The work carried out in this thesis will take this volatility into account.
Tags: Large-scale Distributed Systems, Fault Tolerance, Data Placement

Supervisor: Sebastien Monnet

Co-supervisor: Flavien Vernier

Start of thesis: October 2017

Doctoral school: SISEO "Sciences et Ingénierie des Systèmes, de l'Environnement et des Organisations" (Science and Engineering of Systems, the Environment and Organizations)

Teaching

  • Algorithms and Programming (PEIP, first year)
  • Graphs and Languages (IDU, first year)
  • Algorithmic project (IDU, first year)
  • Object-Oriented Design and Programming (IAI, first year)