Section: New Results

Network Algorithms and Analysis

Bounds on the Cover Time in the Rotor-Router Model

In [23] and  [33] , we consider the rotor-router mechanism, which provides a deterministic alternative to the random walk in undirected graphs. In this model, a set of k identical walkers is deployed in parallel, starting from a chosen subset of nodes, and moving around the graph in synchronous steps. During the process, each node maintains a cyclic ordering of its outgoing arcs, and successively propagates walkers which visit it along its outgoing arcs in round-robin fashion, according to the fixed ordering. We consider the cover time of such a system, i.e., the number of steps after which each node has been visited by at least one walk, regardless of the starting locations of the walks. In the case of k=1, Yanovski et al. (2003) and Bampas et al. (2009) showed that a single walk achieves a cover time of exactly Θ(mD) for any n-node graph with m edges and diameter D, and that the walker explores increasingly large Eulerian subgraphs before eventually stabilizes to a traversal of an Eulerian circuit on the set of all directed edges of the graph.

In [23] , we provide tight bounds on the cover time of k parallel rotor walks in a graph. We show that this cover time is at most Θ(mD/logk) and at least Θ(mD/k) for any graph, which corresponds to a speedup of between Θ(logk) and Θ(k) with respect to the cover time of a single walk. Both of these extremal values of speedup are achieved for some graph classes. Our results hold for up to a polynomially large number of walks, k=O(poly(n)).

In [33] , we perform a case study of cover time of the rotor-router, showing how the cover time depends on k for many important graph classes. We determine the precise asymptotic value of the rotor-router cover time for all values of k for degree-restricted expanders, random graphs, and constant-dimensional tori. For hypercubes, we also resolve the question precisely, except for values of k much larger than n. Our results can be compared to those obtained by Elsässer and Sauerwald (2009) in an analogous study of the cover time of k independent parallel random walks in a graph; for the rotor-router, we obtain tight bounds in a slightly broader spectrum of cases. Our proofs take advantage of a relation which we develop, linking the cover time of the rotor-router to the mixing time of the random walk and the local divergence of a discrete diffusion process on the considered graph.

Web Ranking and Aliveness

In [29] and [30] , we investigate how to efficiently retrieve large portions of alive pages from an old crawl using orderings we called LiveRanks. Our work establishes the possibility of efficiently recovering a significant portion of the alive pages of an old snapshot and advocates for the use of an adaptive sample-based PageRank for obtaining an efficient LiveRank. Additionally, application field is not limited to Web graphs. It can be straightforwardly adapted to any online data with similar linkage enabling crawling, like P2P networks or online social networks.

Wireless Positioning

In [31] , we consider how to construct a low-cost and efficient positioning system. We have proposed a new method called Two-Step Movement (2SM) to estimate the position of Mobile Terminal (MT). By exploiting useful information given by the position change of the device or user movement, this method can minimize the number of Reference Points (RP) required (i.e., only one) in a localization system or navigation service and reduce system implementation cost. Analytical result shows that the user position can be derived, under noisy environment, with an estimation error about 10% of the distance between the RP and MT, or even less.

Content Centric Networking

Today's Internet usage is mostly centered around location-independent services. Because the Internet architecture is host-centric, content or service requests still have to be translated into locations, or the IP address of their hosts. This translation is realized through different technologies, e.g. DNS and HTTP redirection, which are currently implemented at the Application Layer. (ICN) proposes to evolve the current Internet infrastructure by extending the networking layer with name-based primitives.

In [45] , we target the design and implementation of a content router, which is a network entity that implements name-based forwarding, or it can forward packets based on the content name they are addressed to. This work makes three major contributions. First, we propose an algorithm for name-based longest prefix match whose main novelty is the prefix Bloom filter, a Bloom filter variant that exploits the hierarchical nature of content prefixes. Second, a content router design that is compatible with both today's networking protocols and with widely used network equipments. Third, two innovative features that increase the scalability of a content router both in term of forwarding-information-base size and forwarding speed.

In the demonstration [34] held in the ICN conference, we demonstrate a high speed Information-Centric Network in a mobile backhaul setting. In particular, we emulate an information aware data plane and we highlight the significant benefits it provides in terms of both user experience and network provider cost in the backhaul setting. Our setup consists of high-speed ICN devices employed in a down-scaled realistic representation of a mobile backhaul topology, fed with traffic workloads characterized from Orange's mobile network. We compare numerical results activating and de-activating the ICN feature at run-time, showing the main differences between the two approaches. All the devices are implemented in a real high-speed multi-core equipment, and they are connected by means of internal port connections. Traffic is injected using a Traffic Generator which is implemented in the same architecture.

Information Dissemination

Dissemination with Noise or Limited Memory

In [26] , we introduce the study of basic distributed computing problems in the context of noise in communication. We establish tight and almost tight bounds for the rumor spreading problem as well as for the majority-consensus problem.

In [11] , we theoretically study a general model of information sharing within animal groups. We take an algorithmic perspective to identify efficient communication schemes that are, nevertheless, economic in terms of communication, memory and individual internal computation. We present a simple and natural algorithm in which each agent compresses all information it has gathered into a single parameter that represents its confidence in its behavior. Confidence is communicated between agents by means of active signaling. We motivate this model by novel and existing empirical evidences for confidence sharing in animal groups. We rigorously show that this algorithm competes extremely well with the best possible algorithm that operates without any computational constraints. We also show that this algorithm is minimal, in the sense that further reduction in communication may significantly reduce performances. Our proofs rely on the Cramér-Rao bound and on our definition of a Fisher Channel Capacity. We use these concepts to quantify information flows within the group which are then used to obtain lower bounds on collective performance.

Gossip and Rumor Spreading with Flooding

In [2] , we address the flooding problem in dynamic graphs, where flooding is the basic mechanism in which every node becoming aware of an information at step t forwards this information to all its neighbors at all forthcoming steps t'>t. In particular, we show that a technique developed in a previous paper, for analyzing flooding in a Markovian sequence of Erdös-Rényi graphs, is robust enough to be used also in different contexts. We establish this by analyzing flooding in a sequence of graphs drawn independently at random according to a model of random graphs with given expected degree sequence. In the prominent case of power-law degree distributions, we prove that flooding takes almost surely O(logn) steps even if, almost surely, none of the graphs in the sequence is connected. In the general case of graphs with an arbitrary degree sequence, we prove several upper bounds on the flooding time, which depend on specific properties of the degree sequence.

Small-world Networks

In [9] , we study decentralized routing in small-world networks that combine a wide variation in node degrees with a notion of spatial embedding. Specifically, we consider a variant of J. Kleinberg's grid-based small-world model in which (1) the number of long-range edges of each node is not fixed, but is drawn from a power-law probability distribution with exponent parameter α0 and constant mean, and (2) the long-range edges are considered to be bidirectional for the purposes of routing. This model is motivated by empirical observations indicating that several real networks have degrees that follow a power-law distribution. The measured power-law exponent α for these networks is often in the range between 2 and 3. For the small-world model we consider, we show that when 2<α<3 the standard greedy routing algorithm, in which a node forwards the message to its neighbor that is closest to the target in the grid, finishes in an expected number of O(logα-1n·loglogn) steps, for any source–target pair. This is asymptotically smaller than the O(log2n) steps needed in Kleinberg's original model with the same average degree, and approaches O(logn) as α approaches 2. Further, we show that when 0α<2 or α3 the expected number of steps is O(log2n), while for α=2 it is O(log4/3n). We complement these results with lower bounds that match the upper bounds within at most a loglogn factor.

Voting Systems and Path Selection in Networks

In [24] , we apply our theoretical and experimental results on voting systems to a network use case: choosing a path in a network. In our model, nodes have an economical reward or cost for each possible path and they vote to elect the path. We show that the choice of the voting system has an important impact on the manipulability and the economical efficiency of this system. From both points of view, Instant-Runoff Voting gives the best results.