Osnat Mokryn

Propagation and Leader Election in a Multihop Broadcast Environment

The paper addresses the problem of solving classic distributed algorithmic problems under the practical model of Broadcast Communication Networks. Our main result is a new Leader Election algorithm, with O(n) time complexity and O(n · lg(n)) message transmission complexity. Our distributed solution uses a special form of the propagation of information with feedback (PIF) building block tuned to the broadcast media, and a special counting and joining approach for the election procedure phase. The latter is required for achieving the linear time.

Internet resiliency to attacks and failures under BGP policy routing

We investigate the resiliency of the Internet at the Autonomous System (AS) level to failures and attacks, under the real constraint of business agreements between the ASs. The agreements impose policies that govern routing in the AS level, and thus the resulting topology graph is directed, and thus the teachability between Ases is not transitive. We show, using partial views obtained from the Internet, that the Internets resiliency to a deliberate attack is much smaller than previously found, and its teachability is also somewhat lower under random failures. We use different metrics to measure resiliency, and also investigate the effect of added backup connectivity on the resiliency.

On multicast trees: structure and size estimation

This work presents a thorough investigation of the structure of multicast trees cut from the Internet and power-law topologies. Based on both generated topologies and real Internet data, we characterize the structure of such trees and show that they obey the rank-degree power law; that most high degree tree nodes are concentrated in a low diameter neighborhood; and that the sub-tree size also obeys a power law.Our most surprising empirical finding suggests that there is a linear ratio between the number of high degree network nodes, namely nodes whose tree degree is higher than some constant, and the number of leaf nodes in the multicast tree (clients). We also derive this ratio analytically. Based on this finding, we develop the Fast Algorithm, that estimates the number of clients, and show that it converges faster than one round trip delay from the root to a randomly selected client.

An integrated architecture for the scalable delivery of semi-dynamic Web content

The competition on clients attention requires sites to update their content frequently. As a result, a large percentage of Web pages are semi-dynamic, i.e., change quite often and stay static between changes. The cost of maintaining consistency for such pages discourages caching solutions. We suggest here an integrated architecture for the scalable delivery of frequently changing hot pages. Our scheme enables sites to dynamically select whether to cyclically multicast a hot page or to unicast it, and to switch between multicast and unicast mechanisms in a transparent way. Our scheme defines a new protocol, called h.t.t.p.m. In addition, it uses currently deployed protocols, and dynamically directs browsers seeking for a URL to multicast channels, while using existing DNS mechanisms. Thus, we enable sites to deliver content to a growing number of users at less cost and during denial of service attacks, while reducing load on core links. We report simulation results that demonstrate the advantages of the integrated architecture, and its significant impact on server and network load, as well as clients delay.

Tomography of scale-free networks and shortest path trees

In this paper we model the tomography of scale free networks by studying the structure of layers around an arbitrary network node. We find, both analytically and empirically, that the distance distribution of all nodes from a specific network node consists of two regimes. The first is characterized by rapid growth, and the second decays exponentially. We also show that the nodes degree distribution at each layer is a power law with an exponential cut-off. We obtain similar results for the layers surrounding the root of multicast trees cut from such networks, as well as the Internet. All of our results were obtained both analytically and on empirical Interenet data.

Bringing order to BGP: decreasing time and message complexity

The Border Gateway Protocol (BGP), the de facto routing protocol of the internet, generates excessive amount of traffic following changes in the underlying backbone. Previous papers [C. Labovitz, R. Wattenhofer, S. Venkatachary, A. Ahuja, The impact of internet policy and topology on delayed routing convergence, in: Proceedings of the INFOCOM, April 2001; C. Labovitz, A. Ahuja, A. Bose, F. Jahanianitz, Delayed internet routing convergence, in: Sigcomm, September 2000.] show that BGP suffers from high convergence delay and high message complexity after a fail down (detachment) of a network, due to path exploration caused by a limited version of the counting to infinity problem. Surprisingly, we show in this paper that BGP suffers from a high message complexity also after an up event (reattachment of a network). We analyze BGP dynamics data from raw update dumps and show that race conditions cause extensive path exploration that increases the amount of redundant updates. We show, based on these BGP dynamics, that up to 26% of the updates sent during up events are redundant. We also find that the effect of this phenomenon is bigger when the change occurs at the edge of the network. We suggest a minor modification to the waiting rule of BGP that pseudo orders the network and reduces the convergence latency of up events by half and the message complexity from O(DE )t oO(E), where D is the Diameter of the internet and E is the number of connections between ASes. Our simulation results suggest that our modification may improve the convergence messages and time during all events, with the most noted improvement of up to 36% in the number of messages and 81% in time to convergence during up events in Internet like topologies. We show that our results hold also for partial deployment of the modification in only some of the routers.

Sensing clouds: A distributed cooperative target tracking with tiny binary noisy sensors

This paper suggests a novel algorithm for mobile object tracking using wireless sensor networks (WSNs). The paper assumes a future model of WSNs, where a large number of low to medium range inexpensive and noisy sensors are distributed randomly over an area. The distributed algorithm is based on short range communication between neighboring sensors, and is designed to work with very basic low cost binary sensors, that can report only a sensing, not sensing value. Neighboring sensors that sense the object form a cloud around the object which is dynamically updated as the object moves. To save energy on reporting a subset of the cloud, the cloud core, is elected. A trade-off between the accuracy and the core size (namely transmission power) is presented, as well as an extensive simulation study. Our algorithm works well with false negative sensing and up to 10% false positive sensing.

Bringing order to BGP: Decreasing time and message complexity

The Border Gateway Protocol (BGP), the de facto routing protocol of the internet, generates excessive amount of traffic following changes in the underlying backbone. Previous papers [C. Labovitz, R. Wattenhofer, S. Venkatachary, A. Ahuja, The impact of internet policy and topology on delayed routing convergence, in: Proceedings of the INFOCOM, April 2001; C. Labovitz, A. Ahuja, A. Bose, F. Jahanianitz, Delayed internet routing convergence, in: Sigcomm, September 2000.] show that BGP suffers from high convergence delay and high message complexity after a fail down (detachment) of a network, due to path exploration caused by a limited version of the counting to infinity problem. Surprisingly, we show in this paper that BGP suffers from a high message complexity also after an up event (reattachment of a network). We analyze BGP dynamics data from raw update dumps and show that race conditions cause extensive path exploration that increases the amount of redundant updates. We show, based on these BGP dynamics, that up to 26% of the updates sent during up events are redundant. We also find that the effect of this phenomenon is bigger when the change occurs at the edge of the network. We suggest a minor modification to the waiting rule of BGP that pseudo orders the network and reduces the convergence latency of up events by half and the message complexity from O(DE) to O(E), where D is the Diameter of the internet and E is the number of connections between ASes. Our simulation results suggest that our modification may improve the convergence messages and time during all events, with the most noted improvement of up to 36% in the number of messages and 81% in time to convergence during up events in Internet like topologies. We show that our results hold also for partial deployment of the modification in only some of the routers.

The Role of Temporal Trends in Growing Networks.

The rich get richer principle, manifested by the Preferential attachment (PA) mechanism, is widely considered one of the major factors in the growth of real-world networks. PA stipulates that popular nodes are bound to be more attractive than less popular nodes; for example, highly cited papers are more likely to garner further citations. However, it overlooks the transient nature of popularity, which is often governed by trends. Here, we show that in a wide range of real-world networks the recent popularity of a node, i.e., the extent by which it accumulated links recently, significantly influences its attractiveness and ability to accumulate further links. We proceed to model this observation with a natural extension to PA, named Trending Preferential Attachment (TPA), in which edges become less influential as they age. TPA quantitatively parametrizes a fundamental network property, namely the network’s tendency to trends. Through TPA, we find that real-world networks tend to be moderately to highly trendy. Networks are characterized by different susceptibilities to trends, which determine their structure to a large extent. Trendy networks display complex structural traits, such as modular community structure and degree-assortativity, occurring regularly in real-world networks. In summary, this work addresses an inherent trait of complex networks, which greatly affects their growth and structure, and develops a unified model to address its interaction with preferential attachment.

A distributed cooperative target tracking

The paper suggests a novel algorithm for mobile object tracking using wireless sensor networks (WSNs). The paper assumes a future model of WSNs, where a large number of low to medium range inexpensive and noisy sensors are distributed randomly over an area. The distributed algorithm is based on short range communication between neighboring sensors, and is designed to work with very basic low cost binary sensors, that can report only a sensing, not sensing value. Neighboring sensors that sense the object form a cloud around the object which is dynamically updated as the object moves. Our algorithm works well with false negative sensing and up to 10% false positive sensing.