| home |
| Contacts |
- Contact:
- Didier Parigot
- INRIA Sophia Antipolis
- Batiment Fermat, F109
- 2004 Route des Lucioles
- BP 93
- 06902 Sophia Antipolis Cedex
- France
- e-mail: Didier.Parigot@inria.fr
- Tel : (33-4) 4 92 38 50 01
- Fax: (33-4) 4 92 38 76 44
Proposal PhD thesis 2008
Contribution to programmable virtual networks (Overlay Networks)
Environnement
The explosive growth of the Internet gives rise to the possibility of designing large overlay networks and virtual organizations consisting of Internet-connected computers units (peers , or agents ), able to provide a rich functionalityof services that makes use of aggregated computational power, storage, information resources, etc In a nutshell, the objectives of the LogNet research program is to provide adequate notions and definitions of a programmable overlay network computer -- from a desktop distributed calculator (mono application, fixed program) to a programmable distributed overlay computer -- and to propose a precise architecture of a programmable overlay network computer and implement it. The Arigatoni overlay network developed since 2005 in the Mascotte Project Team by Luigi Liquori and Michel Cosnard, and since the 1st January 2008 with Bernard Serpette and Didier Parigot in the LogNet Équipe that I have created at INRIA Sophia Antipolis Méditerraneée, is a structured multi-layer overlay network which provides resource discovery with variable guarantees in a virtual organization where peers can appear, disappear, and self-organize themselves dynamically Arigatoni features essentially two protocols: the resource discovery protocol , RDP, dealing with the process to find and negotiate resources to serve an agent request in its own colony, and the virtual intermittent protocol , VIP, dealing with (un)registrations of agents to colonies. Dealing essentially with resource discovery and peers' intermittence has one important advantage: the complete generality and independence of any offered and requested resource. Arigatoni can fit with various scenarios in the global computing arena, from classical PtP applications (file- or bandwidth-sharing), to new Web applications, to new vehicular-inspired applications, to more sophisticated GRID applications (distributed large computations), until possible, futuristic migration computations , ie transfer of a non-completed local run to another agent, the latter being useful in case of catastrophic scenarios, like fire, terrorist attack, earthquake, etc.
Missions
The objective of this thesis is to explore and develop a execution model on this overlay networks, which will take into account the new constraints of such virtual organization, and especially the particular characteristics of the applications that we will be able to develop on such architectures. Aspects of highly dynamic Architectur, communication peer-to-peer programming and discovery of services will be the topics that will examine this thesis. In particular, we wish to address the following issues programming on a network of recovery:- Establishing a dynamic orchestration that can be taken into account, running, changes the topology of the application (arrival and departure of a resource). This orchestration should be as much as possible decentralized and distributed.
- Direct communication between agents (not centralized communication on a bus integration, Enterprise Service Bus), independent to transport protocols (abstract of the various modes of communication) and lazy communication to better manage the volatile aspect of this organisation.