Maestrois a joint project-team of Inria, Cnrsand the University of Montpellier II (through the Lirmmlaboratory), based in Sophia Antipolis and in Montpellier. It is concerned with the modeling, performance evaluation, optimization and control of discrete-event dynamical systems ( Deds), with a particular emphasis on networks and their applications. The scientific contributions are both theoretical, with the development of new modeling formalisms, and applied with the development of software tools for the performance evaluation of Deds.

Research activities in 2006 have focused on the following issues:

performance evaluation of wireless, mobile, ad hoc and sensor networks, within the Istproject Bionets, a grant from France Telecom R&Dand the support of Cefipra(French-Indian collaboration);

development of online statistical estimators of data flow characteristics with small memory consumptions, supported by the Acigrant Flux;

performance evaluation of peer-to-peer protocols using stochastic fluid models, with the support of France Telecom R&Dand Network of Excellence Euro-Ngi;

analysis of processor-sharing scheduling disciplines, with the support of France Telecom R&Dand Network of Excellence Euro-Ngi;

probabilistic and graph theoretical methods for the Web search engines in the framework of Egide Eco-Netprogram;

singularly perturbed Markov chains and time-average optimal control in the framework of the French-Australian grant Linkage International.

The main mathematical tools and formalisms used in Maestroinclude:

theory of stochastic processes: Markov process, point process, Palm measure, large deviations;

theory of dynamical discrete-event systems: queues, fluid approximation, network calculus;

theory of control and scheduling: dynamic programming, Markov decision process, game theory, deterministic and stochastic scheduling, pathwise comparison;

theory of singular perturbations;

random matrix theory.

Our main application area is networking and in particular, modeling, performance evaluation, optimization and control of protocols and network architectures. It includes:

Internet infrastructure: Tcp, high speed congestion control, voice over Ip, service differentiation, quality of service;

Internet applications: multicast, content distribution systems, peer-to-peer systems, overlay networks, multimedia traffic;

Wireless (cellular, ad hoc, sensor) networks: Wlan, Wimax, Umts, delay tolerant networks, power control, medium access control, transmission rate control, redundancy in source coding, mobility models, coverage, routing.

In 2006, K. Avrachenkov, D. Nemirovsky and N. Osipova have continued to develop a new software ``Web Graph Analyzer'' for the investigation of Web Graph properties. In particular, this software has a crawler with a user-friendly interface, and can efficiently compute the PageRank vector.

Recently, to overcome Aimd(Additive-Increase Multiplicative-Decrease) performance limitations, Westwood Tcpand its Westwood+variant have been proposed. Both of them are based on the new Additive Increase ADaptive Decrease ( Aiadd) paradigm, which leaves unchanged the probing phase of New Reno Tcpand includes an adaptive shrinking phase. In particular, the adaptive decrease used by Westwood Tcpsets the congestion window according to the bandwidth left behind Tcpat time of congestion, which is the end-to-end network available bandwidth. In , E. Altman and his co-authors, including C. Barakat from Inriaproject-team Planete, have analyzed Aiaddprotocols using a two-dimensional recursive equations to obtain the throughput of such protocols. They have first studied the case of i.i.d. Rtts, and then extended their results to the case where Rtts exhibit a Markovian pattern. This research is partly supported by a grant from France Telecom R&Don Internet Traffic (Section ).

In , K. Avrachenkov, in collaboration with L. Finlay and V. Gaitsgory both from the U. of South Australia, has applied the framework of periodic control optimization to study the Aqm– Tcpinteraction. In particular, for a given Tcpversion, the optimal strategy for the Aqmrouter is defined as the solution of a non-linear periodic optimization problem. This solution is obtained via a linear programming approach. It is also shown that depending on the choice of the utility function for the sending rate, the optimal control is either periodic or stationary. A particular attention is paid to a problem with a sigmoidal utility function, in which the evolution of the optimal sending rate exhibits the ``saw-tooth'' behavior of the ``instantaneous'' Tcpsending rate. The sigmoidal utility function corresponds well to audio and video streaming applications. One of the main conclusions of the work in is that an Aqmscheme implementing only two thresholds might be enough. This work is supported by the French-Australian grant Linkage International(Section ).

In , K. Avrachenkov and N. Osipova, jointly with P. Brown ( France Telecom R&D), have derived the conditional expected response time of the two-level processor sharing system when jobs have an hyper-exponential size distribution. This work is a follow-up of , where K. Avrachenkov and his co-authors showed that if the job size distribution has a decreasing hazard rate (which is the case of the hyper-exponential distribution), then the two-level processor sharing scheme outperforms the egalitarian processor sharing scheme. In , it was pointed out that a research topic is the choice of the threshold value, which controls the classification into small and large jobs. In that respect, asymptotic results obtained in will help choosing a value of the threshold that minimizes the total expected response time. This work is carried out within the grant from France Telecom R&Don Internet traffic (Section ).

In the framework of the Fluxproject (supported by the Aci``Masses of pursued the analysis of the problem of recognizing long-lived and short-lived data flows (elephants and mice) using algorithms requiring a (very) limited memory capacity. A statistical estimator for the number of elephants in a data flow has been proposed, and its moments have been analyzed in detail . This estimator has then been extended to more than two flows, and to flows with arbitrary distributions.

Recent advances focus on the application of
*uniform sampling*techniques to this same problem, with the objective to reduce the dependence of the estimator on the detailed stochastic characteristics of the flow. Further research
concerns the improvement of the hashing principle (which is at the heart of the algorithm) towards a better precision with little added memory consumption. Finally, work has begun on the
application of probabilistic counting and sampling techniques to the approximate determination of local characteristics in graphs (such as degrees, girth and clustering coefficients).

During a three-month internship, supervised by S. Alouf and P. Nain, T. Vu has worked on the modeling of the Ieee802.11 medium access control which is based on the Csma/Caprotocol .

Wimaxis a family of standards for wireless metropolitan area networks. Ieee802.16e is one Wimaxstandard providing mobility for users. In , S. Alouf and E. Altman investigate the power save mode defined in the Ieee802.16e standard. To evaluate the transfer delay incurred by the power save mode, and the amount of power it can save, they model the input buffer of a mobile as a single server queue with repeated vacations, in which both idle and busy periods depend on the customers arrival pattern. Current efforts focus on solving the model when packets arrive according to a Lévy input process. This research will be further developed withing the Rnrtgrant Winem(Section ).

This work is carried out within the framework of the Bionetsproject (Section ). The motivation for Bionetscomes from (see http://www.bionets.eu/) ``emerging trends towards pervasive computing and communication environments, where myriads of networked devices with very different features will enhance our five senses, our communication and tool manipulation capabilities. The complexity of such environments will not be far from that of biological organisms, ecosystems, and socioeconomic communities. Traditional communication approaches are ineffective in this context, since they fail to address several new features: a huge number of nodes including low-cost sensing/identifying devices, a wide heterogeneity in node capabilities, high node mobility, the management complexity, the possibility of exploiting spare node resources. Bionetsaims at a novel approach able to address these challenges. Nature and society exhibit many instances of systems in which large populations are able to reach efficient equilibrium states and to develop effective collaboration and survival strategies, able to work in the absence of central control and to exploit local interactions. We seek inspiration from these systems to provide a fully integrated network and service environment that scales to large amounts of heterogeneous devices, and that is able to adapt and evolve in an autonomic way.''

As a first step towards ``bio-inspired networks'', S. Alouf and G. Neglia are investigating the possibility to use an evolutionary approach for epidemic routing in Delay Tolerant Networks.
To route packets over such networks, one may use epidemic routing or one of its variants like probabilistic forwarding, K-hop schemes, or spray-and-wait. These protocols can be captured in a
single framework which allows forwarding schemes to evolve so as to adapt to a changing, and a priori, unknown environment. As mentioned in
, the fundamental components of this framework
are:
(
i)the possibility to share with other nodes a description of the specific forwarding mechanism deployed at each node (this description is called the
*forwarding genotype*),
(
i
i)the possibility for each node to modify its forwarding scheme by applying random changes to it (genotype
*mutations*) or by introducing some characteristics of other nodes genotypes (
*crossover*), and
(
i
i
i)a selection process which fosters the diffusion of the fittest forwarding genotypes, i.e., of those forwarding schemes achieving the best performance in the
specific environment as regards the performance metric chosen (end-to-end delay, power consumption, etc.).

Mobile ad hoc networks (
Manets) often experience route failures and network disconnectivity, especially when the nodes are moving frequently and the network is sparse. M.
Grossglauser and D. Tse have shown (
*Infocom 2001*) that one can take advantage of node mobility in
Manets to increase the network throughput. This can be done by using the so-called ``two-hop relay'' protocol. However, this increase in the
throughout is obtained at the expense of long delays. In order to reduce the (packet) delivery delay A. Al Hanbali, P. Nain, and E. Altman in a series of papers (
,
,
) consider a
Manetthat implements the Multicopy Two-hop Relay (
Mtr) protocol, in which the source is allowed to generate multiple copies of the same packet, unlike in the two-hop relay scheme where the source is
allowed to generate at most one copy. Copies at relay nodes are assumed to have a limited lifetime (time-to-live
Ttl) in the network. Closed-form expressions and asymptotic results when the number of nodes is large are provided for the probability distribution
of the packet delivery delay and for the energy needed to transmit a packet from the source to its destination. Node mobility is captured through the Markovian model developed in
. The results are validated through
simulations for two mobility models, the random waypoint and the random direction mobility models. The performance of the
Mtrprotocol are also compared to that of the epidemic routing protocol, in which both the source and the relay nodes are allowed to generate copies
of a packet.

In , A. Al Hanbali and P. Nain, in collaboration with A. A. Kherani (Indian Institute of Technology, New Delhi), have extended the model and the analysis in to allow for a limited number of copies. This extension is used to identify the number of copies that minimizes the expected delivery delay subject to a constraint on the energy consumption. In this work, they also develop a new model which relaxes the assumption that the inter-meeting times are exponentially distributed. This allows them to evaluate the impact of the exponential assumption on the performance; in particular, they observe that hyper-exponential inter-meeting times yield stochastically larger delivery delays.

This research is carried out within the framework of the Istproject Bionets(Section ).

The work concerns some applications of random matrix theory and unitary random matrix theory to cellular networks. Explicit formulas for performance metrics, depending only on a few meaningful parameters, can be derived in many cases of interest.

In
, N. Bonneau, M. Debbah (
Institut Eurecom) and E. Altman have derived asymptotic results for the performance of a downlink CDMA system with orthogonal spreading and
multi-cell interference when considering an infinite network. The objective is to determine, for a dense and infinite multi-cell network, the optimal distance between base stations. Each
user is equipped with a linear matched filter. The problem is analyzed in the asymptotic regime: very dense networks are considered where the spreading length
Ntends to infinity, the number of users per meter
dtends to infinity but the load per meter
is constant. The theoretical results were shown to be very accurate predictions of the system's behavior in the finite size case. (A shorter version of this work was reported in the
2005
MaestroActivity Report.) This work is carried out within the
Istproject
Bionets(Section
).

The inherent nature of the physical setup and transmission mechanism in wireless ad hoc networks with random channel access, results in correlations between the link metrics of adjacent links, when considering path optimization problems. In , D. Kumar and E. Altman, along with A. A. Kherani (Indian Institute of Technology, New Delhi), identify a special structure inherent to the solution of Dynamic Programming problems arising in such an optimization over paths. According to this structure, the optimal policy tries to equalize the link metrics of adjacent links in a multi-hop route. The authors have validated this structural property with simulations.

In
, D. Kumar and E. Altman study, in
partnership with A. Kumar and V. Ramaiyan (
IIScBangalore), a dense ad hoc network. In a dense multi-hop network of mobile nodes capable of applying adaptive power control, they consider the
problem of finding the optimal hop distance that maximizes a certain throughput measure in bit–meters/sec, subject to average network power constraints. The mobility of nodes is restricted
to a circular periphery area centered at the nominal location of nodes. Using numerical analysis the authors discover that choosing the nearest node as next hop is
*not always*optimal. Optimal throughput performance is also attained at non-trivial hop distances depending on the available average network power. This work was carried out within the
Cefipraproject with India (Section
).

In the above problems, the controls depend on (local) states but they do not influence the state. The control problem can be seen as a repeated game with a random environment. E. Altman, K. Avrachenkov, N. Bonneau, D. S. Menasche, in collaboration with B. Prabhu ( Cwi, Amsterdam) and M. Debbah ( Institut Eurecom), have further studied the more complex case in which the states of the mobiles are controlled as well. These states may represent queue sizes at the mobiles, which may determine quality of service parameters such as delay or loss probabilities , .

During his master internship, supervised by E. Altman and M. Debbah ( Institut Eurecom), A. Suarez-Real has studied uplink power control problems for a continuous power spectrum. Both non-cooperative and adversary situations are considered . This work is supported by Arc Ifany(Section ).

The most popular approach for downlink opportunistic scheduling in cellular networks is the proportional fairness approach. To compute its performance, one usually makes the assumption that the normalized gain between the base station and any mobile is a random variable that does not depend on the location of the mobile (normalized gain refers to the gain at a given point divided by its average value at that point). In K. Khawam and D. Kofman ( EnstParis), in collaboration with E. Altman, propose a more accurate model that defines three zones: (i) zone 0 is the region the closest to the base station – the gain there is assumed to be constant so that the mobiles in this region receive the maximum power; (ii) zone 2 is the farthest from the base station – the maximum power is almost never received there; (iii) zone 1 lies between zones 0 and 2. In various scheduling mechanisms are proposed, in which the relative priority of these zones may vary. For each mechanism, the average throughput in each zone is computed as a function of the control parameters.

In order to compute the capacity of cellular networks, an important parameter is the amount of interference from other cells on any given cell C. This is usually hard to obtain. Often, this amount is approximated to be a constant times the interference generated by cell C. J.-M. Kelif ( France Telecom R&D) and E. Altman propose in an alternative approximation based on replacing the discrete base stations of other cells by a continuous ``fluid'' of base stations. Compared to the discrete cellular network, the approximation is shown to be very accurate. Based on that approximation method, J.-M. Kelif and E. Altman have been able to compute the performance of some scheduling policies that give a priority level to a mobile as a function of its distance from the base station (article in preparation). This work is supported by a grant from France Telecom R&Don Third Generation Mobiles (Section ).

One important issue in ad hoc wireless networks is the characterization of the limiting performance, in terms of both connectivity and coverage. In , E. Altman and D. Miorandi ( Create-Net, Italy) address this issue in a queueing theoretic setting. By using properties of the GI/D/ queueing model, they derive exact results, in the one-dimensional case, for the coverage probability, the node isolation and the connectivity distance for various node placement statistics, in the case of a deterministic channel model. The case of a random channel is investigated with the help of the more general GI/G/ framework. In particular, explicit results are obtained when nodes are distributed according to a Poisson distribution operating in a fading/shadowing environment The approach consists in identifying an equivalence between busy periods in infinite server queues and connectivity ranges in sensor networks. The paper is an extended version of a conference paper reported in the 2004 MaestroActivity Report. This research was supported by the EuroNGISpecific Joint Research Project ``Cellular2'' (Section ).

Surfers on the Internet frequently use search engines to find pages satisfying their queries. However, there are typically hundreds or thousands of relevant pages available on the Web. Thus, listing them in a proper order is a crucial and non-trivial task. PageRank is one of the main criteria according to which Google ranks Web pages. PageRank can be interpreted as a frequency of visiting a Web page by a random surfer and thus it reflects the popularity of a Web page. In , K. Avrachenkov and N. Litvak (U. of Twente) have studied to what extent a page can control its PageRank. Using an asymptotic analysis their provide simple conditions that show whether or not new links result in an increased PageRank for a Web page and its neighbors. Furthermore, they show that there exists an optimal (although impractical) linking strategy. Their main findings are that (1) a Web page benefits from links inside its Web community, and (2) irrelevant links penalize the Web pages and their Web communities.

This research was done in the framework of the Pai Eco-Netgrant (Section ).

As part of a research convention with the startup society Voddnet, A.-E. Baert and A. Jean-Marie have conducted (jointly with V. Boudet, Lirmm, Cnrs/U. of Montpellier II) the realization of a simulation program aimed at optimizing the dimensioning of their contents distribution platform. The program has been realized in collaboration of L. Lacressonière and M. Paniez, from the Master program in Computer Science of the Univ. Montpellier II .

A.-E. Baert and A. Jean-Marie jointly with V. Boudet ( Lirmm, Cnrs/ U. of Montpellier II, France) have worked on data replication and distribution in a particular infrastructure for distributed Video on Demand (VoD). The objective was to minimize the download time of a document under various assumptions. This particular quality of service metric is not commonly encountered in the literature.

In a first work, A.-E. Baert, A. Jean-Marie and their co-authors
have proposed a probabilistic model based
on the
*popularity*distribution of documents. In this setting, minimizing the average download time is a discrete non-linear problem. They showed that this problem can be solved in real
values by Lagrangian optimization. In addition, in a particular case, this problem can be reduced to a knapsack problem. Based on these observations, the authors propose approximation
algorithms validated by simulations.

In a second step, realized with the help of X. Roche (PhD candidate,
Lirmm,
Cnrs/U. of Montpellier II), the authors have studied the optimal data distribution which minimizes the
*variance*of the download time of documents, assuming they have identical popularity. This problem, called ``MINVAR'', can be reduced to a Steiner System Problem (generalizing Steiner
Triple Systems). The authors find an optimality condition for the MINVAR problem. Under appropriate assumptions on documents and the number of hosts, the problem has been solved exactly.
Heuristics and approximation algorithms have been proposed and evaluated
.

During his master internship, supervised by S. Alouf and P. Nain, A. Dandoush has evaluated and compared the performance of two schemes for recovering lost data in a peer-to-peer storage system. The first approach is centralized and relies on a server that recovers multiple losses at once, whereas the second approach is distributed. Through a Markovian analysis and mean fluid approximation, the performance of each scheme has been calculated, in terms of the delivered data lifetime and data availability .

In , S. Alouf, A. Dandoush and P. Nain have extended the model of to account for more elaborate data recovery processes. Numerical computations are provided which assess the impact of each system parameter on the performance of the storage system.

The proportional fairness concept is perhaps the mostly used fairness concept in networking and is widely used. This concept turns out to be a special case of the Nash Bargaining concept from game theory. In , C. Touati, E. Altman and J. Galtier ( Inriaproject-team Mascotteand France Telecom R&D) study properties of this fairness concept as well as ways to compute it efficiency.

E. Altman, N. Bonneau and M. Debbah (Institut Eurécom) study a simplified version of the slotted Alohaprotocol in . Both the cooperative and the non-cooperative approaches are considered. The framework is that of correlated games along with the notion of correlated equilibria. Not only this work applies the notion of correlated equilibria in the context of networking, but it extends it to the multi-criterion case, where each mobile (player) has several objectives. Explicit formulas characterizing the throughput are obtained, and the extent to which coordination could improve transmission in the cooperative and non-cooperative case is analyzed.

G. Neglia has investigated the possibility to apply the framework of network formation games to model the interaction of nodes in sensors networks and in delay tolerant networks. This framework has been employed to study peer-to-peer networks in .

Conjectures have been introduced in game theory in an attempt to model anticipations formed by players in the absence of a complete information. Equilibria reached in a game with
conjectures are called ``conjectural equilibria''. In
(realized in conjunction with M. Tidball (
Inra, Montpellier), A. Jean-Marie has shown that when the idea of conjectures is used in a dynamic learning process, the outcome of the process may
enjoy Pareto-efficiency. This is in relation with the question of the stability of a Cartel in the field of Macroeconomics, raised by Osborne. This last problem is in turn related to the
question of designing
*incentives*(or threats) to regulate non-Nash interactions between economic agents. This issue has been studied in
, where necessary conditions for the
existence of
*Credible Incentive Equilibria*have been obtained, both for static and dynamic games.

Impatience phenomena that cause customers to leave a queue before being served, have been much studied in the queueing literature. These models are useful, in particular, to understand reaction to congestion in call centers or in connecting to a computer network. Most existing models consider impatience due to large queues. In , E. Altman, in collaboration with U. Yechiali (Tel Aviv U.) another phenomenon of impatience: that of customers who find the server on vacation. They have computed performance measures such as the expected number of customers in the system, and the probability that a customer will leave before being served. This research was supported by the EuroNGISpecific Joint Research Project ``Fairness2'' (Section ).

At the joint invitation of O. Boxma (
Eurandom) and R. Núñez Queija (
Cwiand
Tno), K. Avrachenkov and E. Altman, in collaboration with U. Ayesta (
Cwi), wrote a survey on the Discriminatory Processor Sharing (
Dps) discipline
.
Dpsis a multi-class generalization of the egalitarian processor sharing model. In
Dps, all jobs present in the system are served simultaneously at rates controlled by a vector of weights
g_{k}: whenever there are
N_{k}jobs of class
k= 1, ...,
Kpresent in the system, each class-
kjob is served at rate
. The authors hope that the survey will be useful for the networking community, as there are many evidences that the bandwidth sharing in bottleneck nodes in the Internet can be
modeled using
Dpsqueues. This research was supported by the
EuroNGISpecific Joint Research Project ``Fairness2'' (Section
).

Maestrohas pursued its collaboration with France Telecom R&Dat Issy les Moulineaux within a new two-year research grant (Contrat de Recherche Externalisée, Cre). In 2006 a patent has been filed concerning new policies for scheduling Tcpconnections between dedicated and shared channels in Umts . Furthermore, the participants have studied the association problem (which network should a user connect to at any time) within a heterogeneous network environment ( Wlanon one hand, and Umtscellular network on the other hand) . Coordinators of this Creare E. Altman for Maestroand J.-M. Kelif for France Telecom R&D.

Since January 2005 Maestrohas been involved in a two-year research grant ( Cre) with France Telecom R&Din Sophia Antipolis, on the theme ``Internet Traffic Management and Modeling.'' This grant has two major research directions: application of size-based scheduling to Ipnetworks, and analysis of new Tcpversions for high-speed links. In the first direction, in 2006, the participants have derived a simple approximation of the optimal threshold value in Two Level Processor Sharing; in the second research direction, they have proposed the analysis of Tcp-Aqminteraction and carried out a comparison between Tcp New Renoand Tcp Tcp Westwood+.

The topic of this collaboration wit the European Space Agency ( Esa) was the design of efficient online scheduling policies for satellite payloads on launchers. The collaboration was initiated in 2005 and pursued in 2006. It is executed jointly with Ph. Chrétienne and E. Hyon from Lip6, Paris.

The specificity of the problem includes various temporal and feasibility constraints, and a stochastic environment. The collaboration in 2006 concentrated on the extension and test of a discrete-event simulator for the system in consideration .

The mathematical modeling of the problem has led to the Obpresearch proposal that is funded by the Gdr Ro(the French national research initiative on Operations Research), and involves researchers of Maestroand Lip6.

The aim of this project is to develop and to evaluate a software for the tracking of multimedia content published on the Web. This tool will be used by free-lance professional photographers to protect their rights against malicious behaviors. The industrial partner is Canonand the work will be supervised by professional photographers. The kick-off meeting of Spi3-Pro, a one-year project, took place on Dec. 6, 2006.

Within this project, the task of Maestrois to develop a software module which will allow an autonomous and an automatic navigation in the Internet. This work will build on the expertise on Web crawlers developed over the years by Maestro.

This project, called Winem, for WimaxNetwork Engineering and Multihoming, is to be started imminently and will last for 3 years. The project partners are: France Telecom R&D, Get( EnstBretagne and Int), Inria(project-teams Armorand Maestro), Institut Eurecom, Lia(Université d Avignon), and Motorola. It is dedicated to the Ieee802.16e standard for Broadband Wireless Metropolitan Access. S. Alouf is the coordinator for Inria.

Maestrois a member of the Network of Excellence (NoE) EuroNGI on ``Design and Engineering of the Next Generation Internet, Towards Convergent Multi-Service Networks''. E. Altman is the co-coordinator of the work package on ``Control and optimization in telecommunication networks''.

Within this network, Maestrofurther participates in three EuroNGI Specific Joint Research Projects ``JRA.S.07 Cellular2'' and ``JRA.S.11 Fairness2'' (both projects were launched in 2005 and were granted a one-year extension in 2006) and ``JRA.S.23 Delayed'':

Cellular2 (project coordinator: T. Chahed from Get, France) focuses on cross-layer protocol design for wireless networks.

Fairness2 (project coordinator: P. Nain) is devoted to the performance evaluation of fair and efficient scheduling in wired and wireless networks.

Delayed (project coordinator : R. Núñez Queija, Cwiand Tno, The Netherlands) focuses on transfer control with delayed feedback signals.

Within Fairness2 project there have been collaborations between E. Altman and E. Solan (Tel Aviv University) on using game theoretical methods for resources sharing.

Cellular2 project has seen collaborations between T. Chahed ( Int), E. Altman and D. Kumar ( Inria). On Oct. 9, D. Kumar gave a lecture entitled ``Simulation study of a reliable open loop transport for fountain coded traffic in Wlandownlink'' within a Cellular2 meeting organized in Pisa, Italy.

On Nov. 11, K. Avrachenkov participated in a one-day EuroNGI workshop organized within JRA.S.23 Delayed (Eindhoven, The Netherlands) and gave a lecture entitled ``Delay-based congestion control modeled by state-dependent delay differential equations''.

Maestrois a partner of the IstFetEuropean Integrated Project Bionetson ``BIOlogically-inspired autonomic NETworks and Services''. E. Altman is the coordinator of the work package on paradigm collection and foundations.

Bionetsis a project belonging to the IstFetProactive Initiative Program on ``Situated and Autonomic Communication''. There are sixteen partners involved and Maestro, together with Inriaproject-team Oasisand colleagues from Institut Eurecom, is one of them. Bionetsis planned for four years and started on January 1st 2006.

Bionetsspecializes on the design of protocols that will allow evolution of services over a self-organizing wireless network that contains a huge amount of cheap sensors, as well as a limited number of intelligent terminals. The project proposes an inter-disciplinary strategy for designing such networks (called bionets) by using methods and tools from biology, physics, economics. Maestro's task is to collect such tools and to adapt them to Bionets.

Maestro, in partnership with Cnrs(J.-B. Lasserre, H. Frankowska), University of Paris Dauphine (J.-P. Aubin), University of Utrecht (A. Gnedin) and University of South Australia (V. Ejov, J. Filar, L. Finlay, V. Gaitsgory, P. Howlett) participates in a three-year (2005-2007) international cooperation grant, Linkage International, of the Australian Research Council on the subject of ``Singular Perturbations and Multiscale Models in Optimization and Control''.

Since April 2003 Maestrohas been involved in a three-year research grant with A. Kumar and A. Chockalingam ( IIScBangalore) and with V. Borkar ( TataInstitute of Fundamental Research, Mumbai). The coordinators of this project are A. Kumar and E. Altman for the Indian and French sides, respectively. The theme of the cooperation is ``New Strategies for Wireless Communication Networks.'' This cooperation financed the two-year postdoctoral position at Inriaof A. A. Kherani. The program has been extended in 2006 to six more months and has ended on Oct. 31, 2006. Within this program, E. Altman visited the IIScBangalore (Oct. 22–31, 2006) and Maestrohas hosted A. Kumar (June 17–July 1, 2006), A. Chockalingam (July 16–29, 2006) and V. Borkar (Oct. 1–21, 2006).

The Auroraproject is a joint research program funded by the Research Council of Norway and the French Ministry of Foreign Affairs. The participants belong to the following institutions: InriaSophia Antipolis (project-team Maestro; N. Bonneau contributed to this project) and Institut Eurecom, France, UniK - University Graduate Center and University of Oslo, Norway. The objective of this project is to propose advanced optimization tools to design broadband wireless communication systems ( Ofdm, Cdma, Mimo, and ad hoc networks). Within this program, N. Bonneau visited UniK (Oslo; April 12 – May 9, 2006). The complete project plan can be found at Aurora Project website: http://www.unik.no/~arehj/projects/aurora/aurora.html.

Maestro, together with N. Litvak and W. Scheinhardt both from the University of Twente, The Netherlands, is involved in the Pai Eco-Netproject in partnership with V. Dobrynin, A. Garnaev, L. Petrosyan, D. Nemirovsky and S.K. Pham from St. Petersburg State University, Russia, and A. Dudin and V. Klimenok from Belarussian State University, Belarus. Eco-Netis a two-year project (2005-2006) devoted to ``Probabilistic and Graph Theoretical Methods in Search Engines.''

Members of Maestroand the University of Massachusetts ( Umass) at Amherst (D. Towsley) have a long lasting collaboration in the area of performance evaluation and control of networks. Since Sep. 1, 2001, we have been associated with UMassin a five-year Nsf Itrproject on ``QoS in the future Internet''. This project finances visits of members of Maestroto Umass. In 2006, P. Nain visited Umasswithin this framework. The program has ended on Aug. 31, 2006.

Since January 2004 Maestroand Inriaproject-team Oasishave been partners in a four-year collaboration with the University of Los Andes ( Ula), Merida, Venezuela, through a funding of the Ecosprogram. French partners are D. Ros ( Inriaproject-team Armorand EnstBretagne), D. Caromel ( Inriaproject-team Oasis), H. Mounier (University of Paris 11, Orsay), and E. Altman (project coordinator). Our Venezuelan partners are R. Márquez, L. Leon and J. Aguilar from Ula. This year, visits of researchers and students did not include the Maestroproject.

Members of Maestroparticipate in the Arc(Cooperative Research Initiative, sponsored by Inria) Ifany``InFormAtioN theorY: New challenges and new interdisciplinary tools'', whose coordinator is E. Altman. Other Inriaproject-teams involved are Trec, Hipercom, Armorand Temics. Other groups participating are EnstBretagne, University of Avignon, France Telecom R&D, Institut Eurecom, Epfl, University of Cyprus and University of Thessaly. The objective of Arc Ifanyis to bring together methods and tools used in traditional information theory, together with new tools (stochastic geometry, game theory, percolation) so as to come up with new notions of capacity of wireless networks. The website of Arc Ifanyis http://www-sop.inria.fr/maestro/ifany/.

Members of Maestroalso participate in the Arc Coinc, headed by B. Gaujal ( Inriaproject-team Mescal), and aimed at developing computational tools and new applications for what is now known as the ``network calculus'' framework. The website of Arc Coincis http://www-id.imag.fr/Laboratoire/Membres/Gaujal_Bruno/Coinc/coinc.html.

Members of Maestroare involved in two Acis (Incitative Coordinated Actions) sponsored by the Cnrs, Inria, the Ministry of Education and Research and other institutions.

One is the Fluxproject of the Aci``Masses of Data''. Its main topic is the use of probabilistic counting to devise lightweight algorithms for flow classification in networks.

The second one is the Sr2i(Security of Interdomain Routing in the Internet) of the Aci``Security in Informatics'', the topic of which is the reliability and security of the BGP routing protocol.

U. Ayesta ( Laas, Toulouse, France, 04/12/06–04/176/06 and 09/09/06–09/16/06); A. Dudin and V. Klimenok (Belarussian State University, Belarus, 06/12/06-06/22/06); A. Garnaev (St. Petersburg State U., Russia, 06/29/06–07/01/06 and 11/30/06–12/09/06); G. Koole (Vrije University, Amsterdam, The Netherlands, 11/20/06); E. Morozov (Russian Academy of Sciences, Russia, 11/01/06–11/07/06); L. Petrosyan (St. Petersburg State University, Russia, 07/02/06–07/07/06); A. Piunovsky (University of Liverpool, UK, 04/08/06–04/16/06 and 09/10/06–09/17/06).

T. Başar (University of Illinois at Urbana–Champaign, IL, USA, 07/01/06–07/31/06); L. Mason (McGill University, USA, 09/10/06–09/17/06).

E. Solan (Tel Aviv University, Israel, 09/01/06–09/07/06).

A. Chockalingam ( IIScBangalore, India, 07/16/06–07/29/06); H. Kameda (University of Tsukuba, Japan, 08/22/06–09/28/06); A. Kumar ( IISc, India, 06/17/06–07/01/06); V. Borkar ( TataInstitute of Fundamental Research, Mumbai, India, 10/01/06–10/21/06); J. Lui (Chinese University of Hong-Kong, China, 11/20/06 – 11/24/06).

J. Filar (University of South Australia, Adelaïde, Australia, 07/02/06–07/14/06), V. Gaitsgory (University of South Australia, Adelaïde, Australia, 06/26/06–07/08/06).

visited the Indian Institute of Science ( IIScBangalore, India (Oct. 22–31, 2006) and Tel Aviv University, Israel (May 9–23, 2006).

visited Tel Aviv University, Israel (May 15–20, 2006), St. Petersburg State University, Russia (June 5–10, 2006), University of Zielona Gora, Poland (Sep. 16–22, 2006), CWI, The Netherlands (Nov. 7–15, 2006), University of Twente, The Netherlands (Nov. 16-19, 2006).

visited UniK, Oslo (April 12–May 09, 2006) in the frame of Aurora project.

visited the University of Massachusetts, USA (May 15–19, 2006), Vrije University, Amsterdam, The Netherlands (Nov. 13, 2006).

is an Associate Editor of the journals:
*Journal of Economics, Dynamics and Control*(JEDC),
*ACM/Kluwer Wireless Networks*(WINET),
*Communication Networks*(COMNET),
*SIAM Journal on Control and Optimization*(SICON), and
*Journal of Discrete Event Dynamic Systems*(JDEDS).

is an Associate Editor for
*RAIRO Operations Research*and
*Performance Evaluation*.

is an Associate Editor of
*IEEE/ACM Transactions on Networking*,
*Performance Evaluation*and
*Operations Research Letters*.

was a program committee member of the
*First International Conference on Performance Evaluation Methodologies and Tools (Valuetools 2006*), Oct. 11–13, 2006, Pisa, Italy.

was a program committee member of the following conferences:

*International Workshop on Protocols for Fast Long-Distance Networks (PFLDNET 2006)*, Nara, Japan, Feb. 2–3, 2006,

*Workshop on Modeling and Optimization in Mobile, Ad Hoc and Wireless Networks (WiOpt 2006)*, Boston, MA, USA, April 3–7, 2006 (he was also a member of the steering
committee),

*IEEE International Workshop on Quality of Service (IWQoS 2006)*Yale University, New Haven, CT, USA, June 19-21, 2006,

*European Wireless Conference (EW 2006)*, Athens, Greece, 2-5, April 2006,

*Wireless On-demand Network Systems (WONS 2006)*, Les Menuires, France, Jan. 18–20th, 2006,

*12th International Symposium on Dynamic Games and Applications*, Polytech, U. Nice Sophia Antipolis, July 2006, France,

*International Conference on AD-HOC Networks and Wireless (AdHocNow)*, Ottawa, Canada, Aug. 17–19, 2006,

*IEEE Infocom 2006*, Barcelona, Spain, April 23-27, 2006,

*IFIP Networking 2006*, Coimbra, Portugal, May 15–9, 2006.

was a program committee member of the following conferences:

*The 6th International Conference on Next Generation Teletraffic and Wired/Wireless Advanced Networking (NEW2AN 2006)*, St. Petersburg, Russia, May 29 – June 2, 2006,

*The 14th IEEE International Conference on Networks (ICON2006)*Singapore, Sep. 13–15, 2006,

*IEEE Globecom 2006*, San Francisco, CA, USA, Nov. 27 – Dec. 1, 2006.

was a program committee member of the following conferences:

*MAMA 2006*(held in conjunction with Sigmetrics/Performance 2006), Saint-Malo, France, June 26–30, 2006,

*3rd European Performance Evaluation Workshop 2005 (EPEW 2006)*, Budapest, Hungary, June 21–22, 2006,

*3rd International Conference on the Quantitative Evaluation of SysTems (QEST 2006)*, Univ. of California at Riverside, CA, USA, Sep. 11–14, 2006,

*Asian Internet Engineering Conference 2006 (AINTEC 2006)*, Bangkok, Thailand, Nov. 28–30, 2006,

*21st International Symposium on Computer and Information Sciences (ISCIS 2006)*, Istanbul, Turkey, Nov. 1–3, 2006.

was a program committee member of the following conferences:

*IEEE Infocom 2006*, Barcelona, Spain, April 23–27, 2006,

*21st International Symposium on Computer and Information Sciences (ISCIS 2006)*, Istanbul, Turkey, Nov. 1–3, 2006),

*MAMA 2006*(held in conjunction with Sigmetrics/Performance 2006), Saint-Malo, France, June 26–30, 2006.

organized and chaired the program committee of the
*First International Conference on Performance Evaluation Methodologies and Tools (Valuetools 2006*), Oct. 11-13, 2006, Pisa, Italy.

He also co-organized (together with H. Levy and I. Eliazar) a three-day workshop in Shefayim, Israel, at the event of the retirement of U. Yechiali (Tel Aviv University).

E. Altman is now working on the preparation of a Special Issue of the journal
*Probability In the Engineering Sciences*(PEIS) based on selected papers from the workshop.

was a member of the organizing committee of the
*12th International Symposium on Dynamic Games and Applications*(July 3–6, 2006, Sophia Antipolis).

She organized the kick-off meeting of the Arc Ifany(March 29–30, 2006, Enst, Paris).

participated as a reviewer in the PhD thesis committee of Kinda Khawam, EnstParis (Nov. 14, 2006) and in the HdR thesis committee of Bruno Tuffin (April 10, 2006). He also participated as a thesis co-advisor in the PhD thesis committee of Ahmad Al Hanbali (U. of Nice Sophia Antipolis, Nov. 20, 2006).

participated as thesis co-advisor in the PhD thesis committee of Gregory Miller (MAI, Moscow, Russia, Dec. 22, 2006).

participated as a reviewer in the PhD thesis committee of P. Chambreuil (Tech. U. Belfort-Montbéliard and U. Franche-Comté, Jun. 23, 2006), and in the PhD thesis committee of M. Bouklit (U. Montpellier II, July 5, 2006).

In 2006 he has been appointed member of the Best Thesis Award Committee of SPECIF, the French Society for Education and Research in Computer Science.

participated as thesis co-advisor in the PhD thesis committee of Ahmad Al Hanbali (U. of Nice Sophia Antipolis, Nov. 20, 2006), and in the PhD thesis committee of Menno Dobber (Vrije University, Amsterdam, The Netherlands, May 13, 2006). P. Nain was a member of the Best Thesis Award Committee of Telecom Valley (Provence-Alpes–Côte d'Azur Region).

The following PhD theses were defended in 2006:

(Advisors: E. Altman and P. Nain) on November 20, 2006 (committee: G. Koole, J. Lui (reviewers), H. Afifi, E. Altman, E. Biersack, W. Dabbous, P. Nain).

(Advisors: K. Avrachenkov and A. Pankov) on December 22, 2006 (committee: E. Y. Rubinovich, E. A. Rudenko (reviewers), A. Pankov, K. Avrachenkov, A. Borisov). ``These en co-tutelle'' between the University of Nice Sophia Antipolis, France, and Moscow Aviation Institute, Russia.

is a member of the Doctoral Committee of InriaSophia Antipolis.

is Head of the Apr(Algorithms and Performance of Networks) project-team of the LirmmLaboratory, a joint research unit of CNRS and the University of Montpellier II.

is co-chair of the Department of Computer Science of the LirmmLaboratory, Montpellier (80 permanent staff).

is a member of the Recruiting Committee (Commission de Spécialistes) in Computer Science at the University of Montpellier II.

is a member of the Steering Committee of the GDR RO, a national research initiative on Operations Research sponsored by the Cnrs.

is a member of the management of the Research Unit of InriaSophia Antipolis.

is Head of project-team Maestro.

is Vice-Head of the Project-Committee of the Research Unit of InriaSophia Antipolis.

is a member of the Evaluation Committee of Inria.

is in charge of the Master program on ``Networks and Distributed Systems'' at the University of Nice Sophia Antipolis.

is a member of the Steering Committee of the Anr(National Research Agency) research program in Telecommunications.

is a member of the Scientific Committee of the Graduate School of Information and Communication Sciences (école doctorale Stic) of the University of Nice Sophia Antipolis.

is treasurer of Ifip Wg7.3 on ``Computer System Modeling''.

A. A. Kherani, R. El-Azouzi and E. Altman
were awarded the
**Best Paper Award**of
*IFIP Networking 2006*(May 2006, Coimbra, Portugal).

E. Altman, A Jean-Marie and P. Nain are (elected) members of Ifip Wg7.3 on ``Computer System Modeling''.

participated in the course on ``Probability and Statistics'' in the Engineering Program at Polytech' Nice-Sophia (52H).

taught a course on the Network Simulator (ns-2) within the Master Program on ``Networks and Distributed Systems'' at the University of Nice Sophia Antipolis (24H).

taught courses in the Master in Informatics of the University of Montpellier II on Performance Evaluation (6H), ``Quality of Service in Networks'' (27H), Random Discrete Structures (27H) and Metrology and Quality of Services (68H). She participated in the course on ``Communication and Networks'' of the Master in Computer Science of the University of Montpellier II (30H).

participated in the course on ``Dynamics and Algorithmics of Networks'' of the Master in Computer Science MPRI (Paris 6/ENS/École Polytechnique, 6H). He taught a course on Random Discrete Structures in the Master in Informatics of the University of Montpellier II (20H).

taught a course on the Performance Evaluation of Networks in the Master Program on ``Networks and Distributed Systems'' at the University of Nice Sophia Antipolis (24H).

presented a paper at
*Infocom 2006*(Barcelona, Spain, April 23–29) and at the
*First International Conference on Performance Evaluation Methodologies and Tools (Valuetools 2006*) (Oct. 11–13, 2006, Pisa, Italy). He also gave a poster presentation at the
*IEEE International Workshop on Quality of Service (IWQoS 2006)*(Yale University, New Haven, CT, USA, June 19-21, 2006).

attended (as an author of a paper)
*IFIP Networking 2006*(Coimbra, Portugal, May 15–19, 2006). She also attended the
*
Euro-Ngi
workshop on Stochastic Performance Models for Resource Allocation in Communication Systems(Amsterdam, Netherlands, Nov. 8–10, 2006).*

was an invited plenary speaker at the
*12th International Symposium on Dynamic Games and Applications*(Sophia Antipolis, France, July 2006) and gave a presentation at
*Infocom 2006*(Barcelona, Spain, April 23–29, 2006). He was also an invited speaker in the
*Second Madrid Conference on Queueing Theory (MCQT)*(July 2006).

was an invited plenary speaker at the
*First International Conference for Young Researchers in Computer Science, Control, Electrical Engineering and Telecommunications (ICYR'06)*(Sep. 18-20, Zielona Gora, Poland). He has
also given presentations at the
*Workshop in honour of Uri Yechiali*, (Shefayim, Israel, May 17–19, 2006), the
*6th International Conference on Next Generation Teletraffic and Wired/Wireless Advanced Networking (NEW2AN 2006)*(St.Petersburg, Russia, May 29 – June 2, 2006), and the EuroNGI
workshop organized within JRA.S.23 Delayed (Eindhoven, The Netherlands, Nov. 11, 2006). He also attended the
*
Euro-Ngi
workshop on Stochastic Performance Models for Resource Allocation in Communication Systems(Amsterdam, Netherlands, Nov. 8–10, 2006).*

gave a presentation at
*IFIP Networking 2006*(Coimbra, Portugal, May 15–19, 2006).

presented two papers at the
*4th International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt 2006*) (April 3–7, 2006, Boston, USA) and one paper at
*IFIP Networking 2006*(May 15-19, 2006, Coimbra, Portugal). He also gave an invited talk at Euro-Ngi Cellular2 meeting (Oct. 9, 2006, Pisa, Italy).

was an invited speaker at the
*EuroNGI Workshop on Stochastic Performance Models for Resource Allocation in Communication Systems*(Amsterdam, The Netherlands, Nov. 8–10, 2006). He attended (as an author of a paper)
*Infocom 2006*(Barcelona, Spain, April 23–29, 2006).

gave a presentation at
*ACM SIGCOMM workshop on Challenged Networks*(Pisa, Italy, Sep. 15, 2006).

has given presentations at the
*First International Conference on Performance Evaluation Methodologies and Tools (ValueTools'06)*(Pisa, Italy, Oct. 11–13, 2006),
*EuroNGI Workshop on Stochastic Performance Models for Resource Allocation in Communication Systems*(Amsterdam, The Netherlands, Nov. 8–10, 2006).