Our research activities range from theoretical bounds to practical protocols and implementations for large-scale distributed dynamic systems. The target applications range from Internet-based applications to wireless autonomic networks. We focus our research on two main areas: resource management and dissemination. We believe that such services are basic building blocks of many distributed applications. We also examine these services in two networking contexts: Internet and wireless sensors. These two classes of applications, although exhibiting very different behaviors and constraints, clearly require scalable solutions.
To achieve this ambitious goal, we tackle the issues both along the theoretical and practical sides of scalable distributed computing and ASAP is organized along the following themes:
Models and abstractions: dealing with dynamics,
Resource management in large-scale dynamic systems,
Peer-to-peer wireless autonomic networked systems.
In a system composed of
nprocesses where at most
tcan exhibit a Byzantine behavior, it is known since early eighties that
tneed to be smaller then a third of the total number of process to make the Byzantine consensus problem decidable. Moreover, it has been proved that the minimum number of communication
steps needed is
t+ 1in the worst case. Yet, this protocol is extremely costly in terms of the size of messages and local computation (this protocol is called EIG for Exponential
Information Gathering). So far, lowering this cost has led to consider smaller values of
t(such as
t<
n/4) and an increased latency (
2
t+ 2steps). One protocol exist with a reasonable (polynomial) cost with the initial number of steps but is extremely complex. The goal of this work is to design a
simple algorithm that is as simple and resilient as the EIG protocol but with a quadratic number of steps.
In contrast to the client/server approach, peer-to-peer systems originally consider nodes to have equal capabilities and roles. File sharing applications, for example, have showed the power of heterogeneity by identifying nodes with extra capabilities and making them play different roles. For the more general purpose of taking benefit of heterogeneity in dynamic systems, we investigated the solutions to the distributed slicing problem by extending our previous work on the topic. This work has been done in collaboration with Vincent Gramoli (EPFL & University of Neuchatel, Switzerland), Ymir Vigfusson, Ken Birman and Robbert van Renesse (Cornell University). More specifically, we proposed Sliver a fast converging algorithm achieving accurate slicing.
This work has been done in collaboration with Rachid Guerraoui and Maxime Monod (EPFL, Switzerland). The emergence of social networks provides a framework for polling a community easily by the mean of peer-to-peer techniques. Polling is not as critical as voting as the accuracy on the tally is less important. Yet, it must provide similar properties to electronic voting, such as voters privacy, fairness, probabilistic accuracy, ...The core idea of the work is to build a decentralized protocol without cryptography ensuring this properties with high probability by the mean of peer-to-peer deterrent power : every action in the protocol may be subject to verification by peers. Ensuring that any malicious action is detected with probability one or at least close to one, we increase the accuracy of the tally by limiting the proportion of peers misbehaving. Privacy is ensured probabilistically by using peers as proxies for emitting ballots making vote recovery impossible for reasonable proportion of malicious nodes.
We applied the
condition-basedapproach to the
k-set agreement problem. In a brief annoucement in PODC'08, we introduced the definition of a framework that allows defining conditions suited to the
-set agreement problem. More precisely, a condition is defined as a set of input vectors such that each of its input vectors can be seen as “encoding”
values, namely, the values that can be decided from that vector. A condition is characterized by the parameters
t,
, and a parameter denoted
dsuch that the greater
d+
, the least constraining the condition (i.e., it includes more and more input
vectors when
d+
increases, and there is a condition that includes all the input vectors when
d+
>
t). The conditions characterized by the triple of parameters
t,
dand
define the class of conditions denoted
,
0
d
t,
1
n-1. The properties of the sets
are investigated, and it is shown that they have a lattice structure.
In ICDCS'08
, our contribution is a generic synchronous
k-set agreement algorithm based on a condition
, i.e., a condition suited to the
-set agreement problem, for
k. This algorithm requires at most
rounds when the input vector belongs to
C, and
rounds otherwise. (Interestingly, this algorithm includes as particular cases the classical synchronous
k-set agreement algorithm that requires
rounds (case
d=
tand
= 1), and the synchronous consensus condition-based algorithm that terminates
in
d+ 1rounds when the input vector belongs to the condition, and in
t+ 1rounds otherwise (case
k=
= 1).)
We studied autonomous mobile robots moving on a finite anonymous graph and focuse on the
Constrained Perpetual Graph Explorationproblem (
CPGE). That problem requires each robot to perpetually visit all the vertices of the graph, in such a way that no vertex hosts more than one robot at a time, and each edge is traversed by
at most one robot at a time. This work has been done in cooperation with Roberto Baldoni professor in Roma and Alessia Milani post-doc in Madrid. François Bonnet spent one week in Roma to
work on this subject and he also presented this work during a seminar ALPAGE (February'08) and as an invited speaker in LIP6 (November'08).
In an Information Processing Letters paper
, we provided an upper bound
kon the number of robots that can be placed in the graph while keeping
CPGEsolvability. In a short paper in OPODIS'08, The second contribution restricts. We propose algorithms that solve the
CPGEproblem up to a certain number
kof robots depending on a given parameter
that captures the view of each robot, on partial anonymous grid (a finite grid with possibly missing vertices or edges).
We focus on the problem of providing a reliable, geographically localised storage service. This problem can be abstracted as providing a virtual shared register, i.e. an object equipped with read and write functionalities, attached to a particular location and maintained by a set of mobile nodes. Such an abstraction can be seen as a basic building block to be used for the implementation of higher level services like a localised warning system or a localised black board.
The preliminary results of our work are published as short papers in EDCC'08, DEBS'08, and OPODIS'08 conferences. We specified and implemented such a shared object for non concurrent writes and multiple readers, then we describe a possible extension to concurrent writes. We also provide insights on how this kind of objects could be useful to program location-aware applications. This work is a collaboration between people from Laas in Toulouse, people from La Sapienza in Roma, and people from IRISA in Rennes as a mini-project in the European Network of Excellence ReSIST. In this context, François spent few days in Roma, and in Toulouse twice.
Population protocols provide theoretical foundations for mobile tiny device networks in which global behavior emerges from a set of simple interactions between anonymous agents. The works in this area mostly focus on studying the computational power of the model. Results hold as long as a fair scheduler, ensuring that all reachable system states are endlessly reached, governs the interactions between nodes. This assumption is crucial to ensure that the protocols eventually converge.
We studied for the first time the impact of the agents mobility model on the convergence speed of population protocols. We perform our study by considering several mobility models traditionally used in the ad-hoc network community. We propose an augmented population protocol model where each edge of the interaction graph is weighted, representing the probability of two agents to interact. This models the behaviour of the scheduler with respect to various mobility models. We empirically show that mobility models do have a significant impact on the convergence speed of the protocols. In fact, we observe that the uniform distribution always provides the best convergence time. Such a model is representative of the well-known random-way point model used to evaluate most of mobile ad-hoc network protocols. We then formally proved that a uniform distribution of weights provides the lowest bound of average convergence speed for any population protocol.
This work started on September 1st 2008 and is the topic of the ERC Starting Grant led by Anne-Marie Kermarrec. Traditional search engines are performing extremely well but do hardly encompass alternative and very dynamic sources of information such as blogs, p2p file sharing system instant messaging as well as some modern forms of content distribution frameworks. This is mainly due to their lack of adaptivity to dynamics and their not taking into account correlations between contents and users preferences. They are also limited by their reliance upon centralized indexing. Typically, corporate pages are visited frequently while individual information may be visited rarely: the individual is at a disadvantage. This reveals a striking evidence that complementary and novel fully decentralized alternatives to traditional search engines are now required to capture the dynamic, collaborative and heterogeneous nature of the digital universe as well as leverage individual preferences and social affinities. At the heart of GOSSPLE lies the idea that leveraging the virtual communities of users and the social dimension of the network in a dynamic manner is the key towards reaching the many areas of the Internet applications that are not covered by current search engines. GOSSPLE stems from the observation that indexing the data is not really the main issue, collaborative and dynamic navigation is. GOSSPLE is an autonomic, completely decentralized self-organized network, in which participants get connected to the virtual sub-network, i.e., an overlay network, matching the operations they are performing.
We initiated GOSSPLE in the context of user-centric search. In this context, we worked on the metrics to assess the distance between users in arge-scale systems. Collaborative filtering algorithms are information retrieval techniques used to detect interest patterns of users for items. They are useful to predict the interest of a user for a given item. Those technique are already deployed in many applications, but they only work in a centralized context and are therefore limited by storage and calculation time constraints. We developed fully decentralized algorithms to automatically group users whose interests are similar and provide each user with a list of items that match his profile. We focused both on the quality of the proposed items and their diversity. Our approach was validated through simulations on data from the music community website lastFM and eDonkey file sharing logs. We are currently investigating more complex tag-based systems. Vivien Quéma (SARDES) and Rachid Gerraoui (EPFL) are currently collaborating with us on this topic.
Geographic routing is a hot topic in the wireless sensor network area, it consists to route a message from any nodes to the closer node of the destination coordinates. In this topic we improve previous works particularly for complex topology. Our first contribution was to propose a routing protocol which achieves optimal routing performance with a message and memory overhead per node independent of the network size.
M. Bertier served in organization committee for the following conferences: Algotel 2008 Rencontres Francophones sur les aspects Algorithmiques de Télécommunications, Saint-Malo, France, May 2008. DEBS'08 2nd Int'l Conference on Distributed Event-Based Systems, Rome, Italy, Jul 2008. He served also in program committes of the following conferences:
Algotel 2008 Rencontres Francophones sur les aspects Algorithmiques de Télécommunications, Saint-Malo, France, May 2008. CFIP'08: Colloque francophone sur l'ingénierie des protocoles, Arc, France, 2008. CFSE'08: Conférence Française en Systèmes d'Exploitation, Fribourg, 2008 IWSOS 2008, the third International Workshop on Self-Organizing Systems, Vienna, Austria, December 2008
Gossip-based protocols are simple, robust and scalable and have been consistently applied in many distributed, mostly wired, settings. Most validation in this area has been so far empirical and there is a clear lack of a theoretical counterpart clearly defining what can and cannot be computed with gossip-based protocols. Population protocols, on the other hand, provide a clear and formal model for mobile sensor networks, capturing their power and limitations.
We established a correlation between population and gossip-based protocols. Studying the equivalence between them, we propose a classification of gossip-based protocols, based on the nature of the underlying peer sampling service. First, we show that the class of gossip protocols, where each node relies on an arbitrary sample, is equivalent to population protocols. Second, we show that gossip-based protocols, relying on a more powerful peer sampling providing peers using a clearly identified set of other peers, are equivalent to community protocols, a variant of population protocols. Leveraging the resemblances between these areas enables to provide a theoretical framework for distributed systems where global behaviours emerge from a set of local interactions, both in wired and wireless settings.
This work has been done in collaboration with Etienne Rivière and Pascal felber (university of neuchatel, Switzerland) and Spyros voulagris (ETH Zurich). Gossip-based protocols are well known to provide a simple, scalable and extremely robust way to disseminate messages in large-scale systems. We proposed a generic efficient push-pull dissemination protocol, PULP, which combines the best of both worlds. PULP exploits the reliability and efficiency of push approaches, while limiting redundant messages and therefore imposing a low overhead, as pull protocols do. PULP relies on the dissemination of a stream of messages to achieve this: by pushing some messages, PULP enables an efficient pulling of other messages, which in turn help with the dissemination of the stream. We have deployed PULP both on a cluster and on PlanetLab. Our results demonstrate that PULP achieves an appealing trade-off between robustness, redundancy, and delays.
This work has been done in collaboration with Maarten Van Steen and Vivek Rai (Vrije Universiteit Amsterdam). Video-on-demand (VoD) is increasingly attractive internet application allowing a client to begins to play a movie stored on a server almost instantaneously. The missing pieces of the video are downloaded in a sequential order while playing the downloaded pieces. Peer-to-peer swarming has proved to be a very efficient paradigm for file distribution even with a very low proportion of seeds provided that all the peers actually contribute to the system by uploading pieces to others. This problem is critical in VoD applications since peers must achieve their deadlines, and thus rely on the other peers, to provide a smooth playback to the user. The incentives used by Bit Torrent, referred as Tit for tat are not usable in VoD downloads due to the sequential nature of VoD downloads and since any two peers may not have mutual interest.
Our work consist in building an efficient Incentive-compatible peer-to-peer Video-on-Demand system by imposing a loose structure on the peer sets : grouping nodes whose positions in the video are close in groups and build a linked list of groups to ensure feedinggroups by more advanced ones in exchange of most advanced pieces injected in less advanced groups. Using an appropriate pieces seeding/feeding strategy the full swarm comes down to several very efficient independent swarms fed by each others, the seed being appealed only by the most advanced group. The structured VoD swarming protocol outperforms existing protocols based on random graphs in terms of throughput, goodput (i.e., playback rate) and memory usage (by keeping a limited memory of past pieces instead of storing the entire video).
In small-world networks, each peer is connected to its closest neighbors in the network topology, as well as to additional long-range contact(s), also called shortcut(s). In 2000,
Kleinberg provided asymptotic bounds on the routing performance and showed that greedy routing in a
npeer small-world network, performs in
steps when the distance to shortcuts is chosen uniformly at random, and in
O(log
2n)when the distance to shortcuts is chosen according to a harmonic distribution in a
d-dimensional mesh. Following our theoretical analysis in 2007 in which we refined the routing complexity measure for small-world networks, we are now working on the practical
evaluation of a Kleinberg based peer sampling. More specfically, we are currently working on the design of gossip-based protocols providing a good approximation of Kleinberg-like
small-world topologies. To this end, we bias the peer sampling protocol so that the sampling is biases toward a Kleinberg distribution. We showed that such a biased peer sampling impact
only a little the properties of the resulting graph such as the average path length, the clustering coefficient and the in-degree distribution.
This work has been done in collaboration with Vivien Quéma (CNRS Grenoble), Maxime Monod and rachid Guerraoui (EPFL). Gossip-based information dissemination protocols are considered easy to deploy, scalable and resilient to network dynamics. Load-balancing is inherent in these protocols as the dissemination work is evenly spread among all nodes. Yet, large-scale distributed systems are usually heterogeneous with respect to network capabilities such as bandwidth. In practice, a blind load-balancing strategy might significantly hamper the performance of the gossip dissemination.
These observation, collected in a white paper presented at LADIS 2008 as an invited contribution , have led to the development of HEAP, HEterogeneity-Aware gossip Protocol, where nodes dynamically adapt their contribution to the gossip dissemination according to their bandwidth capabilities. Using a continuous, itself gossip-based, approximation of relative bandwidth capabilities, HEAP dynamically leverages the most capable nodes by increasing their fanout, while decreasing by the same proportion those of less capable nodes. HEAP preserves the simplicity and proactive (churn adaptation) nature of gossip, while significantly improving its effectiveness. We extensively evaluated HEAP in the context of a video streaming application on a 270 PlanetLab node testbed. Our results show that HEAP significantly improves the perceived quality of the streaming over standard gossip protocols, especially when the stream rate is close to the average available bandwidth.
Gossip-based dissemination reaches all nodes with high probability. However, some peers may not receive all the data. In order to enhance the performance of gossip-based dissemination protocol, fountain codes can be used to recover the missing data. We propose a protocol that combine a gossip-based dissemination protocol and fountain codes. The gossip protocol builds a structured split-graph overlay to split the peers between encoders or forwarders. Forwarders become encoders as soon as they have received the whole content so they can start encoding. In order to benefit even further from encoders, we deliberately bias the dissemination process through the gossip protocol so that forwarders become earlier potential encoders. We implemented this protocol on the PlanetLab testbed. By simulation in PeerSim, we assess that this protocol outperform a simple protocol that do not introduce a bias or do not update the overlay to match the peers'role.
DHTs and RPS are two very important services used by many P2P applications. DHT based systems provide a efficient lookup functionality while peer sampling services enable to build and maintain connectivity in the presence of high churn and offer a lot of flexibility. In addition many applications, in particular search applications, may want to combine them to leverage the benefits of both approaches. We propose to integrate a p2p structured overlay and gossip-based overlay sampling service to build a DHT on the fly. The small-world nature of DHTs based systems is captured by the use of two cohabiting gossip-based protocols: SRLS maintains a set of short links (numerically speaking) and LRLS selects long-range links.
Our approach has several advantages: It can start from arbitrarily connected overlay and build the system on the fly as nodes gossip. It is very simple and unlike other DHT protocols has no special mechanism to handle churn and even extreme cases such as catastrophic failures. The set of long-range which are changing continuously provide a very useful feature of distributing the routing load evenly across the system. The indegree of the nodes are well balanced.
This work has been done in collaboration with Amy Murphy (Trento, Italy). Trees are fundamental structures for data dissemination in large-scale network scenarios. However, their inherent fragility has led researchers to rely on more redundant mesh topologies in the presence of churn or other highly dynamic settings. In this work , we identified a novel protocol that directly and efficiently maintains a tree overlay in the presence of churn. The protocol simultaneously achieves other beneficial properties such as limiting the maximum node degree, minimizing the extent of the tree topology changes resulting from failures, and limiting the number of nodes affected by each topology change. Applicability to a range of distributed applications is discussed and results are evaluated through extensive simulation and a PlanetLab deployment.
The storage capacity of computers has increased a lot in the past years: in the meantime, final users have started using this storage for important personal data, with the democratization of digital cameras, and professional data with the rise of telecomputing. Backuping all this data has become a new challenge for peer-to-peer systems, since these users are connected most of the time, often with large unused storage capacity on their disks, and unfortunately seldom take the time to properly save these important data.
Anne-Marie Kermarrec and Fabrice Le Fessant are currently designing a platform for a collaborative backup system, and this problem tackles a large set of problems: making the backup resilient to the large number of failures characterizing peer-to-peer networks, choosing where to backup the data, designing the protocols to place and retrieve the data from the network, while ensuring secrecy/privacy of the data. The prototype, currently developed by Fabrice Le Fessant within the Peerple open-source project, uses both a structured overlay, to localize stored data during restoration, and an unstructured overlay, to query for storage availability among neighbors. Contrary to most peer-to-peer backup systems, files are not stored separately on the overlay network, but gathered in volumes, encrypted using strong cryptography for privacy, and replicated using Reed-Solomon coding, to ensure availability even in the presence of high failure rates at a minimal extra storage cost. This work is done in collaboration with Laurent Viennot from the Gangproject-team, INRIAParis - Rocquencourt.
Many of distributed systems are dynamic in that participants may leave or (re-)join the system unpredictably. The key parameter of such systems is called the churn, i.e. the dynamism intensity. To face dynamism, existing applications traditionally use costly recovery mechanisms that overestimate the churn. As a consequence, the communication cost employed to recover despite unbounded failures prevents these applications from scaling as well as they could. To avoid the need for upper bounds on churn rate, decentralized estimation schemes provide a way for every system node to self adapt to network characteristics; to provide an estimation, recent approaches assume an underlying structured network ( DHT) and the fact that churn events are uniformly distributed all over the network. Unfortunately, this strongly restricts their use.
We present an algorithm that is, as far as we know, the first distributed method for monitoring churn in arbitrary networks, with also arbitrary node departure patterns. This method relies on a constant number of message exchanges between each node and its neighbors, and thus scales well. It uses a predetermined period of monitoring. During this period, the nodes monitor the dynamic events occurring in the system, and locally compute a partial information about the churn. When this period ends, dedicated nodes aggregate the computed information to obtain a global estimate of the churn. Results have been published in 2008 in conferences CFSE and PODC .
This work is done in collaboration with Bruno Séricola (DYONISOS). We propose a distributed algorithm to assess the connectivity quality of a network, be it physical or logical. In large
complex networks, some nodes may play a vital role due to their position (
e.g.for routing or network reliability). Assessing global properties of a graph, as importance of nodes, usually involves lots of communications; doing so while keeping the overhead
low is an open challenge. To that end,
centralitynotions have been introduced by researchers to rank the nodes as a function of their importance in the topology. Some of these techniques are based on computing the ratio
of shortest-paths that pass through any graph node. This approach has a limitation as nodes “close” from the shortest-paths do not get a better score than any other arbitrary ones. To avoid
this drawback, physician Newman proposed a centralized measure of
betweenness centralitybased on random walks: counting for each node the number of visits of a random walk travelling from a node
ito a target node
j, and then averaging this value by all graph source/target pairs. Yet this approach relies on the full knowledge of the graph for each system node, as the random walk target node
should be known by advance; this is not an option in large-scale networks. We propose a distributed solution that relies on a single random walk travelling in the graph; each node only
needs to be aware of its topological neighbors to forward the walk. A preliminary paper has been published at DISC conference in 2008
.
Tracking peer availability in a peer to peer network is of utmost importance for many collaborative applications. For instance, such information is invaluable for identifying the most stable peers or group of peers with similar uptime characteristics. However, as many applications tend to reward the most stable peers, there is a clear incentive for peers to try to appear more available than their real availability. We developed a scalable and lightweight protocol that enables nodes to measure the peer availability in the presence of such selfish peers. In our protocol, which is called Pacemaker , each peer is in charge of maintaining proofs of its own availability over time by collecting pulses disseminated by a trusted entity using asymmetric cryptographic signatures. Essentially, using these pulses, peers gain the ability to challenge other peers and verify that their real uptime matches the advertised one. Simulation results show that our protocol provides accurate availability measures even in the presence of selfish peers. Furthermore, our results are verified by experiments in Planetlab, which also illustrates the deployability of Pacemaker in real networks.
We worked on the design, correctness, and analysis of SONDe, a simple fully decentralized object deployment algorithm for highly requested systems. Given an object (service or data), SONDe provides a node with a constant upper bound (h) on the number of logical hops to access an object holder (provider), thus making tunable and predictable the communication latency between a node and any provider. In addition, SONDe is able to dynamically adapt the number of providers to reflect load variations experienced in localized portions of the system. Each node individually decides to be a provider, based on the observation of its h-hops neighborhood. We show theoretically that SONDe self-stabilizes and provides an independent-dominating set of providers. Finally simulation results, conducted over different network topologies, demonstrate the efficiency of the approach and confirm the theoretical analysis. This research has been published at EDCC conference in 2008 .
This work has been done in collaboration with Roberto Baldoni and Leonardo Querzoni (University La sapienza, Rome). We consider a set of anonymous moving objects to be tracked in a binary sensor network, we studied the problem of associating deterministically a track revealed by the sensor network with the trajectory of an unique anonymous object, namely the Multiple Object Tracking and Identification(MOTI) problem. In our model, the network is represented by a sparse connected graph where each vertex represents a binary sensor and there is an edge between two sensors if an object can pass from one sensed region to another one without activating any other remaining sensor. The difficulty of MOTI lies in the fact that trajectories of two or more objects can be so close (track merging) that the corresponding tracks on the sensor network can no longer be distinguished, thus confusing the deterministic association between an object trajectory and a track.
We first showed that MOTI cannot be solved on a general graph of ideal binary sensors even by an omniscient external observer if all the objects can freely move on the graph. Then, we restricted the graph, on the object movements or both, to make MOTI problem always solvable. In the case of absence of the omniscient observer, we showed that our results can lead to the definition of distributed algorithms that are able to detect when the system is in a state where MOTI becomes unsolvable.
We conducted studies in information dissemination in mobile networks where a contemporaneous path may never exist between two nodes (a source and a destination) in the network, known as delay-tolerant networks (DTNs). These networks are further challenged by strict resource constraints (e.g., memory, CPU and energy limitations). Our goal in this domain was to support intelligent and adaptive forwarding, which allows a good trade-off between reliability and resource-efficiency. In particular, the main challenge for adaptive mechanisms in DTNs is to adjust forwarding decisions on the fly and still ensure reachability, even when the network is not always connected and disconnections are hard to predict. We then design a new protocol, called Seeker, which empowers nodes with the ability to estimate favorable contact opportunitiesby taking advantage of any information they can locally infer. Hence, nodes are able to adapt and self-organize in dynamic environments with minimal control overhead. This work has been performed in collaboration with the researcher Roy Friedman (from Technion - Israel Institute of Technology) and is currently in submission. The performed research work was also inserted in the context of the RnrtSVP project, which has finished in September 2008. Related to this described research a study about: data dissemination using gossip algorithms in wireless networks was published in the ACM SIGOPS Operating Systems Review (ACM OSR) and routing in wireless self-organizing networks was published in a chapter of the book Adaptation and Cross Layer Design in Wireless Networks, Taylor and Francis Group .
This work has been done in collaboration with Eric Fleury (ENS LYON) Most Wireless Sensor Networks (WSN) applications require connected-coverage of the monitored zone. We consider an application where external entities such as human beings ask the network (through any of its sensor) to sense at a given location in the Region of Interest (ROI) and we call the fulfilment of such a request routing. Using mobile sensors, we propose a routing algorithm independent from the communication devices (i.e., which does not assume a disc model for the communication graph) which provides, in addition to satisfying sensing requests, transparently self-deployment of the network in the ROI. The sensors are not aware of the ROI but only satisfy request either by moving, forwarding or fulfilling sensing requests without any knowledge of the ROI (size, shape, boundaries,...). We prove formally that the network converges to a triangular lattice providing greedy-connected coverage of the region of interest (i.e., once the system has converge no move is needed to fulfil a request anymore) as the number of requests increases. The triangular lattice requires a number of nodes within a constant factor of the optimal connected-coverage deployment. We provide simulations results which confirm the theoretical study and demonstrate the efficiency of the protocol. In addition, we consider practical matters making the algorithm usable in real-life environment. This work has been done in collaboration with Eric Fleury (ENS Lyon / INRIA Rhônes-Alpes).
The motivation behind this research work comes from the lack in the literature, of an autonomous system able (1) to permanently evolve and self-organize under dynamic changing conditions (due either to the environment or technological issues), and (2) to provide various networking functionalities over the same underlying support system. We argue that networks must not only scale in size but also in functionality. In particular, we observe that the related proposals are designed focusing on various different segment of network functionality: network slicing, virtual coordinates, data aggregation, load balancing, etc.
In general, due to the way they are designed, adding a not previously envisaged new feature requires the whole network reconfiguration and/or the provision of new node capabilities. These requirements can be, however, invalidated or deferred if the network is deployed in an area of difficult access. Our answer to those demands was an autonomous system able not only to be adaptable to environment changing conditions, but also, that provides variety to network functionalities. Thus, we have designed an autonomous and lightweight self-organizing networked system that, by imposing a bounded overhead to wireless devices, constructs a base network structure for supporting network functionalities, commonly required in WSNs. We have only exploited local connectivity information and per-neighbor communication. This work has been performed in collaboration with the researcher Roy Friedman (from Technion - Israel Institute of Technology) and is currently in submission. The research related to this subject has been published: at the PODC 2008 (as a brief announcement) and in CFIP 2008 conferences, at the lettre bimestrielle Réeseaux et Télécoms , and recently at the ICDCN conference to be held in January 2009.
The research activities in this domain concern a typical environment consisting in a large number of sensor nodes deployed to collect data or events in a specified geographic area and in a mobile sink moving over the region to collect monitored data. The main challenge in such a context is to safely store collected data such that they can be retrieved later.
We have been working on how to make the sensed monitored data available to the mobile sink in a robust, adaptive, and efficient way. In this context, We have proposed, in collaboration
with the researchers Artur Ziviani (from the National Laboratory for Scientific Computing (LNCC), Brazil) and Roy Friedman (from the Technion - Israel Institute of Technology) an efficient
data dissemination approach (in terms of overhead and representativeness) to allow a mobile sink to gather a representative view of the monitored region covered by
nsensor nodes by visiting
any
mnodes, where
m<<
n. This work is currently in submission. This research also constituted the M2R internship subject of Mubashir Rehmani, student of SUPELEC under the supervision of
Aline C. Viana. Since September 2008, Mubashir is a Phd student at the UPMC Univ Paris 06, under the supervision of Aline C. Viana and the supervision of the Prof. Serge Fdida. Related to
this research a study about data aggregation in wireless sensor networks was also published as a chapter of the book Ad Hoc Networks: New Research, Nova Science Publisher, Inc.
.
The vast literature on the wireless sensor research community contains many valuable proposals for managing energy consumption, the most important factor that determines sensor lifetime. The goal of this work is to extend the network lifetime. We aim at determining good energy-efficient routes in the network by using the energy level of nodes as a criterion to select good links in the route. The estimation of node energy level in the network is not a trivial task, since in a wireless radio communication the remaining energy of a node can be affected by many factors: its data transmission, a data reception, and the interference caused by a closer data transmission. Interesting researches have been facing this requirement by focusing on the extension of the entire network lifetime: either by switching between node states (active, sleep), or by using energy efficient routing.
In collaboration with Khaldoun Al Agha and the PhD student Joseph Rahme both from the LRI/Universite Paris-Sud, our first contribution in this context was the proposal of cost functions to choose energy efficient routes. The originality of the proposed cost functions lies in their completely decentralized and adaptive behavior in considering energy consumption, remaining energy of nodes, and the number of transmissions a node can make before its energy depletion. The gotten results were published in the Springer Annals of Telecommunication Journal and in the IFIP IHN 2007 conference.
A collaboration with Marcelo Dias de Amorim from CNRS/LIP6 focus on applications of wireless sensor networks that require periodic readings. This means that these readings should be
performed following some predefined parameter
fminthat denotes the minimum frequency at which the whole target area must be sensed. An interesting solution for this problem is to use mobile sensors that move around in order to cover
multiple target regions. Previous related works addressed
whymobility is useful in WSNs. In contrast, we also consider mobile sensor network, but focus on
howsensors should move, in order to guarantee the coverage of all targets in the network in a timely and efficient way. To answer this issue, we proposed a mobility strategy that
consists in making nodes follow a Hilbert space-filling curve and use opportunistic contacts to reduce the data delivery's delay. The first results on this work were published at the ACM
HeterSanet 2008 workshop
.
This work deals with malicious behaviors in the context of sensor networks. Such a behavior can be due to an adversary that has some sensors under control or more generally to a problem of the sensor itself. Effectively, as sensors are small devices that are industrially built, many of them may be defective. Moreover, it is known that when a sensor is running out of energy, it can enter a state where it abnormally behaves. Malicious behaviors in sensor networks less hard to handle as the power of the adversary is lower. Indeed a sensor has a limited energy. The more it is active the less it will survive and thus even its computation power is bounded. In the case of a sensor network with static sensors, we try to build secured links between sensors. The objective is to avoid the case of an adversary that collects the whole information exchanged among the sensors.
We studies the message efficiency problem of Byzantine fault tolerant broadcast in multi-hop wireless networks. We tries to find the most message or energy efficient ways to accomplish a reliable broadcast despite malicious nodes that may alter the message or cause collisions. We show that the network's ability to tolerate malicious node depends on nodes' message bound, the malicious node density and the topology form by the nodes which have already receive the message. Initial results were presented as short paper at DISC 2008.