|
|
Version
française
Scientific bases: Controling and optimizing the IP network
In the beginning, the sole purpose of the Internet was to transmit information packets from one end of the network to the other. The Internet was thus not designed for such real time applications as IP telephone or video. Such data is encapsulated so as to be transported by the network and go through the same channels, via the TCP protocol. The need for increasingly reliable equipment at all points of the network cannot be ignored any longer. Routers, servers, caches and links must thus be totally mastered from one end of the chain to the other. However, the defects encountered (packet loss, congestion) cannot be overcome without a profound revision of the architecture. It therefore becomes necessary to model such networks, in order to try and predict their behavior and install them optimally. Thus, at INRIA, a whole range of techniques, models and algorithms, based on probability theory or random geometry, are used to attempt to meet this challenge, by evaluating network performance and analyzing the traffic.
Research on these themes is carried out by projects HIPERCOM,
FRACTALES,
RAP, TREC,
and PREVAL
among others. Project TREC and the Alcatel Network Strategy Group are collaborating
on a joint strategic operation that aims to highlight the characteristics
of the traffic and to design models used to optimize the functioning of large
IP networks, in particular ADSL networks.
Project MISTRAL
also developed a tool called WAGON that is capable of emulating different
types of Internet traffic, to study server behavior. WAGON is being used in
the framework of the VHTD national project.
However, questions of optimization are just as crucial in networks. Projects DREAM and SIGMA2 are monitoring and diagnosing network malfunctionings using preestablished models. The objective of project ARMOR, a joint project of INRIA, CNRS, the University of Rennes 1 and the ENST Bretagne (National Superior School of Telecommunications), is to set up new architectures that take into account these numerous constraints, using tools capable of working on the flows (quality of service), or of analyzing packet behavior. The team is also very interested in the new protocols that are integrated into the future network generations, such as IPv6 (new generation Internet), or MPLS (Multilabel Protocol Switching) a very popular quality of service protocol.
The field of digital communications also presents many challenges, either for the radio layer, optical transmissions or multiuser shared access. This field is currently underrepresented at INRIA, given the needs. In the field of coding and algorithmic protection of information, project CODES is developing techniques based on discrete mathematics and algorithmics. Projects SIGMA2 and TEMICS are working in close collaboration on joint source/channel protection and space-time codes for multiuser access, and are studying the "turbo" methods by considering them as probabilistic models on graphs.
Projects involved: RAP, ARMOR, DREAM, TREC, FRACTALES, SIGMA2, MISTRAL,
HIPERCOM, CODES and PREVAL
Contact: François Baccelli
, project TREC
Software infrastructure for networks: The services of tomorrow
Network infrastructure would be nothing without the accompanying software architecture and new services. This is especially true for upcoming domestic innovations. Many INRIA research projects are working on the new protocols and architectures that will make up future applications.
The implementation, evolution and optimization of such protocols is thus one of INRIA's main goals.
For example, project RESEDAS is focusing on three types of networks : new generation Internet (IPv6), programmable networks and ad hoc mobile networks. The team is striving to create synergies between these distinct platforms by working on a unique experimental platform whose purpose is to integrate new services, such as service level management, routing protocols and network component monitoring functions.
For their part, the RESO research team is trying to propose new solutions in terms of protocols, services and software for the networks that will constitute the future Internet and infrastructures (grid computing). The essential goal is to improve high speed performance and develop new services and protocols.
Project SARDES is interested in the development of software infrastructures (operating systems, middleware) that are adjustable to large scale, heterogeneous and dynamic distributed environments. In particular, project SARDES is working on the design of distributed infrastructures that can be administered and dynamically reconfigured, the development of monitoring and configuration management functions for distributed systems, their exploitation to build grid high performance application servers and configurable infrastructures for mobile multimedia applications. Project SARDES is also one of the principal promoters and contributors to ObjectWeb, an open international consortium whose objective is the development of open software middleware technology.
Finally, the goal of project
VASY is to design industrial quality methods and software that detect
errors as early as possible and increase confidance in the systems studied.
In particular, VASY is actively contributing to the development of the CADP
toolbox for protocol engineering. The main activities of the project are compilation,
rapid prototyping, simulation, partial or exhaustive verification and testing.
Since mobility is one of the major challenges in the years to come, INRIA
is also working on software gourmands en énergie) en sont les maîtres
mots.
Project ARLES, for instance, is setting out to define new software architectures for distributed systems on mobile sites, in order to offer users an acceptable quality of service while taking into account the relatively limited resources of some of the user terminals, the variations of available bandwidth and the global system load. Project ARLES is also contributing to projects that concern ambient intelligence. A consortium AIR-D (Ambient Intelligence Research and Development) was founded in collaboration with Philips and Thomson multimedia, with the goal of creating an intelligent domestic environment. Such devices as sensors, screens and vocal interfaces can be used to directly interact with objects in the day-to-day surroundings. The primary targets of this research are intelligent homes and cars. Miniaturization and autonomy (modest energy needs) are the key words there.
Project ACES for its part is studying a new approach to the use of pocket computers based on neighborhood interactions. The idea is to set up direct communications between physically close mobile devices without resorting to any fixed infrastructure. As soon as exchanges are possible, the two pocket computers become W-PDAs (for wireless PDA) and form a "spontaneous" information system. The team is studying the processes needed to implement such spontaneous information systems. Such wireless communication technologies as Bluetooth (which is already on the market) will make such systems possible.
Projects concerned: ARLES, ACES, SARDES, RESEDAS, RESO, VASY
Contact: Jean-Bernard Stefani,
project SARDES
Mobile communications: The major challenge of the decade
Mobility is certainly one of the most pressing issues of the decade, in terms
of cell phones, portable computers and electronic appointment books. Succeeding
in connecting any mobile terminal to the Internet anywhere any time remains
a challenge of the most difficult kind.
INRIA is carrying out a lot of research with this goal in mind, to propose
new models, protocols and applications that meet the mobility needs, taking
into account all the constraints that it entails.
Such is the case of ad hoc networks, or next generation mobile networks.
These neighborhood networks are innovative and have wind in their sails. However,
they are still very controversial for economic and political reasons. In practice,
an ad hoc network is a mobile network that does not use any infrastructure,
as opposed to the cellular networks that are already in place. The mobile
nodes relay the information to one another. This type of network is interesting
because they can be up to 1000 times faster than the telephone or 10 times
faster than ADSL. This is why such networks involve important economic issues.
The telephone operators who invested lots of money in UMTS licences (3rd generation
phones) see this technology as a harsh contender. Such networks are of great
interest to the military to set up mobile tactical networks adapted to troop
movements, as well as companies that can extend their Ethernet networks in
situations where wiring is difficult.
Within INRIA, the HIPERCOM
team is devoted to developing these new standards. The team is the originator
of the HIPERLAN standard that was adopted in 1996 by the ETSI (European Telecommunication
Standard Institute) for wireless data transmission. A wireless routing protocol
called OSLR was defined in the framework of this research and is currently
being standardized by the IETF (Internet Engineering Task Force).
Project PLANETE
is working in parallel to improve the performance of wireless networks using
the 802.11 standard defined by the IETF and a competitor of HIPERLAN.
Project ACES for its part is interested in spontaneous information systems based on ad hoc networks, in which mobile devices exchange information depending on their physical proximity. Two mobile computers, each equipped with a communication module and within communication range of each other, can directly exchange messages. As soon as exchanges become possible, the two computers involved form an information system that has the special feature of being spontaneous.
Still in the framework of innovative networks, the ARES team is intent on deploying services on such architectures, either in ad hoc or wireless LAN mode, or by combining the two modes. The major directions of research are the interoperability of systems and protocols, as well as the optimization of radio, network and system resources when a service is deployed and used.
Mobility is however extremely demanding and other constraints come along
on top. In effect, mobility undoubtably implies displacements. A machine sends
its correspondents a temporary IP address during each of its journeys, either
local or global, of whatever extent. A study shows that 69% of a user’s
displacements are local. In this context, project PLANETE is working on a
solution that would separate global mobility from local mobility, according
to two different protocols. The load of the Internet is thus reduced and the
risks of attack by signaling message hacking are minimized. The solution implemented
is also compatible with the Mobile IP protocol of the IETF.
Finally, implementing new protocols also constitutes an important issue. Project
TREC also comes in the analysis of mobile infrastructures. The project is
collaborating with France Telecom on devising mathematical models for cellular
communications based on CDMA (especially UMTS) and on the design of new load
control protocols for such networks.
Projects concerned: HIPERCOM, PLANETE, ACES, ARES and TREC.
Contact: Philippe Jacquet,
project HIPERCOM
Internet: Toward an integration of transportation media
Even though it is spread throughout the globe, the Internet is nonetheless very heterogeneous. All sorts of media-satellite, ADSL, cable, mobile and so on-with different throughputs, transport the data through the network. It would thus be ideal to design infrastructures that make this heterogeneity transparent. INRIA is working on this goal by devising protocols and applications that are not linked to a particular technology and can work everywhere, in spite of the disparity of the equipment that make up the network. Naturally, this is not an easy task since each transmission medium has its own characteristics. However, the dominance of the IP protocol, which can no longer be ignored, now makes possible what was not possible with other technologies.
From this perspective, project PLANETE has demonstrated its inventiveness by designing the UDLR protocol, that favors traffic on asymmetrical links, such as for example, transmissions from satellites using the digital broadcasting standard DVB. This solution is interesting because it has an important effect on the development of Internet by satellite.
A startup called UDCast was founded in
June 2000, whose purpose is to promote this standard. UDLR is now validated
by the whole international community since it is standardized by the IETF
(Internet Engineering Task Force).
This research on the Internet protocol are carried in parallel with research
by the ARMOR, MISTRAL and TREC teams, that work on understanding and defining
protocols and procedures for a better quality of service on the Internet.
Projects concerned: PLANETE, ARMOR, MISTRAL, TREC.
Contact: Walid Dabbous, project
PLANETE
Optical communications at the heart of the infrastructure
Optical technology is increasingly present at the core of the infrastructure, especially in the case of very high speed networks. These networks do not however withstand any anomaly or error. The quality of the network is indeed crucial. In this context, project SIGMA2 intends to monitor this quality of service using a method that is totally transparent, i.e. independent of the throughput or the data coding. A patent was registered jointly at the INPI.
In the case of optical networks, it is also crucial to minimize equipment costs. Project MASCOTTE is applying itself to large network dimensioning in order to reduce the costs. The team has also developed a modular network dimensioning software.
Being at the heart of the infrastucture, fiber optics are becoming the most studied medium. Projects MIAOU and ESTIME are using mathematical methods to model, identify and control optical systems such as fibers or the entirely optical regenerators of optical signals that have to reconstitute and amplify a degraded incoming signal.
Finally, project CONGE is developing tools and methods for the control of polarization in fiber optics.
Projects concerned: MASCOTTE, CONGE, SIGMA2, ESTIME, MIAOU
Contact: Albert
Benveniste, project SIGMA2
