Master d’informatique 2008 2009
srsar
Proposition de thèse

Title : Services Discovery and stabilization in large-scale networks

Location : MIS Lab./Amiens

Contact : Franck Petit, LiP6 (à partir de septembre), Franck.Petit@ens-lyon.fr Vincent Villain, MIS, Vincent.Villain@u-picardie.fr

Description : The emergence of Petascale architectures, the evolution of grid computing, and the accession of peer-to-peer (P2P) systems increase a lot the number of potentially resources. However, right now, existing infrastructures and access rules do not allow to fully take advantage of these resources.

One key idea of Project SPADES is to propose a non-intrusive but highly dynamic environment able to take advantage to potentially accessible resources without disturbing their native use. For instance, in a grid computing environment, volatile resources must be requested via batch schedulers to reservation mechanisms which are limited in time or susceptible to preemption (best-effort mode). It is then important to maintain an up-to-date, distributed data structure on the system state. Such a monitoring must allow users discovering proposed services.

From a general point of view, the subject of this thesis discusses the feasibility of a software solution for service discovery that can withstand the high dynamic nature of system components (services, resources, communication links, etc.).

A few numbers of works on peer-to-peer systems address self-stabilization. The concept of self-stabilization is a general technique to design a system to recover by itself from an arbitrary configuration. It confers the system to tolerate arbitrary transient faults. It is also suitable for networks where nodes or links are added and removed, as in peer-to-peer networks.

Nevertheless, self-stabilization cannot “mask” the disturbances made by such additions or deletions. During the disturbances, the system can return arbitrary responses to user requests. Ideally, when a user would request a service, he should receive an accurate response, regardless the system state.

A snap-stabilizing guarantees that the system always maintains the desirable behavior. This concept could help in the design of dependable solutions, i.e., ensuring reliable response to the user requests.

The PhD thesis agenda is mainly twofold. First, it will consist in establishing conditions to consider self-stabilizing solutions to implement some of the basic tasks used in the implementation of a discovery service architecture, such as synchronization and leader election. Second, the student will study the feasibility of snap-stabilizing solutions to these problems.

Support : Project SPADES, ANR (http://graal.ens-lyon.fr/SPADES)