Nos partenaires



Accueil du site > English > Departments > Architecture, Systems, Networks > SEPIA Team


Système d’exploitation, systèmes répartis, de l’intergiciel à l’architecture

Responsable : Jean-Marc Pierson , Daniel Hagimont


The general problematic of the SEPIA team is centered on the efficient conception and usage of distributed operating systems, more precisely for clusters, grids and today clouds.

A system is characterized by the resources it manages, the abstraction it provides and its objectives. These concepts are studied in SEPIA from the hardware to large-scale middleware.

Resource management is complex due to the heterogeneity of the resources. These can be of different types, varying in size, quantity, characteristics (life cycle for instance), implying that a single management can’t be adapted to all cases : In clusters, grids and clouds, resources vary extremely.

Abstractions provided to application developers are highly dependant on the application domain. For middleware several dedicated frameworks exist for different application domains.

Objectives of these systems are numerous. They allow for defining evaluation criteria of designed resource management policies (i.e. performance optimization or energy savings) and pertinence criteria of the abstractions provided to the developers (i.e. genericity, or adaptability for specific domain).

Three axis have been identified in the team :

Axis 1 : Mobile process management by system virtualization

The objective of this axis is to study techniques based on virtual machines in order to provide the abstraction of a set of processes able to migrate between physical machines. Virtual machines provide transparent migration when distribution and communication are involved.

We study the interest of this abstraction (mobile processes) for optimizing the placement of processes so as to reduce resource consumption (number of machines) or to reduce communication costs (process collocation).


Axis 2 : Resource usage optimization : energy and performances

The objective of this axis is to determine the parameters allowing for an efficient usage of resources in terms of energy consumption and performances. Then to use these parameters for two main directions :

  • (1) dynamic placement of processes in order to reduce the energetic impact of data and computing centres ;
  • (2) optimal placement of data as a function of the target architecture. This resource management needs a resource monitoring and to interact with the resource provider. Tradeoffs or consensus have to be obtained in the system thanks to dedicated protocols.


Axis3 : Autonomic system management system

The objective of this axis is to study the principles of the conception of an autonomic system and its application in several domains. When TUNe provides high level languages to facilitate the definition of management policies (deployment, reconfiguration, …), we saw that to handle new administrative tasks or new application domains forced us to define new more adapted languages. Therefore we explore the conception of a system allowing to integrate easily new languages as a function of the considered domain.

Several application domains are studied at the moment : embedded systems for avionics and automotive characterized by different software components ; large scale scientific computing ; energy management in clusters, grids and clouds (link with axis 2) ; management of deployment, configuration and reconfiguration of the grid file system Visage.

Cette rubrique ne contient aucun article.