## RaWMS -- Random Walk based Lightweight Membership Service for Wireless Ad Hoc Networks (2008)

### Cached

### Download Links

- [www.ee.technion.ac.il]
- [webee.technion.ac.il]
- [static.googleusercontent.com]
- [www.cs.technion.ac.il]
- [www.cs.technion.ac.il]
- [www.csl.mtu.edu]
- [www.ee.technion.ac.il]
- [webee.technion.ac.il]
- [www.cs.technion.ac.il]
- [www.cs.technion.ac.il]
- [www.cs.technion.ac.il]
- [csn.cs.technion.ac.il]
- [www.cs.technion.ac.il]

Citations: | 29 - 7 self |

### BibTeX

@MISC{Bar-Yossef08rawms--,

author = {Ziv Bar-Yossef and Roy Friedman and Gabriel Kliot},

title = {RaWMS -- Random Walk based Lightweight Membership Service for Wireless Ad Hoc Networks},

year = {2008}

}

### OpenURL

### Abstract

This paper presents RaWMS, a novel lightweight random membership service for ad hoc networks. The service provides each node with a partial uniformly chosen view of network nodes. Such a membership service is useful, e.g., in data dissemination algorithms, lookup and discovery services, peer sampling services, and complete membership construction. The design of RaWMS is based on a novel reverse random walk (RW) sampling technique. The paper includes a formal analysis of both the reverse RW sampling technique and RaWMS and verifies it through a detailed simulation study. In addition, RaWMS is compared both analytically and by simulations with a number of other known methods such as flooding and gossip-based techniques.

### Citations

5034 | Matrix Analysis - Horn, Johnson - 1985 |

2526 | Malts. Dynamic source routing in ad hoc wireless networks - Johnson, A - 1996 |

2399 | The capacity of wireless networks
- Gupta, Kumar
(Show Context)
Citation Context ... our purposes. Specifically, the G2(n, r) graph, also known as the Unit Disk graph, is often used to model the network connectivity graph of 2-dimensional wireless ad hoc networks and sensor networks =-=[19, 20]-=-. See Appendix A for a formal description of the model. We assume that nodes do not know their position and we do not use any geographic knowledge in our algorithms. Each node has a unique identifier ... |

2133 | Collective dynamics of ‘small-world’ networks - Watts, Strogatz - 1998 |

2012 | On the evolution of random graphs - Erdös, Rényi - 1960 |

1917 | Randomized Algorithms
- Motwani, Raghavan
- 1995
(Show Context)
Citation Context ...the mixing time). There are a number of methods for obtaining a loose upper bound on the network size, e.g., [15, 36]. Once we have such a loose upper bound, we can use the birthday paradox principle =-=[32]-=- to obtain a much tighter bound. This is by keeping track of the average time between publications of the same identifier. We have shown that according to the reverse sampling technique, every time so... |

1285 | Untraceable Electronic Mail, Return Addresses, and Digital Pseudonyms - Chaum - 1981 |

766 | A Measure of Asymptotic Efficiency for Tests of a Hypothesis Based on the Sum of Observations - Chernoff - 1952 |

671 | The small-world phenomenon: An algorithmic perspective - Kleinberg - 2000 |

568 | Search and Replication in Unstructured Peer-to-Peer Networks
- Lv, Cao, et al.
- 2002
(Show Context)
Citation Context ...ear in [28] and [21]. The idea of using a “maximum-degree” random walk in order to reach a uniform limit distribution on the state space has been used before in a number of contexts [4, 7]. Lv et al. =-=[30]-=- propose to use simulated random walks for searching in unstructured peer-to-peer networks. They report that such a search is preferable to the standard practice of searching by flooding. The suitabil... |

421 | Critical power for asymptotic connectivity in wireless networks, Stochastic Analysis, Control, Optimization and Applications: a volume in honor of
- Gupta, Kumar
- 1998
(Show Context)
Citation Context ... our purposes. Specifically, the G2(n, r) graph, also known as the Unit Disk graph, is often used to model the network connectivity graph of 2-dimensional wireless ad hoc networks and sensor networks =-=[19, 20]-=-. See Appendix A for a formal description of the model. We assume that nodes do not know their position and we do not use any geographic knowledge in our algorithms. Each node has a unique identifier ... |

338 | Group communication specifications: A comprehensive study
- Chockler, Keidar, et al.
(Show Context)
Citation Context ...variety of other services and applications in ad hoc networks. A membership service provides each node with a view regarding who are the other nodes in the network. In traditional membership services =-=[10]-=-, the view of each process approximates the entire membership. Moreover, views must be consistent, and changes to views must be coordinated among all their members. This complete and strongly consiste... |

325 | Tarzan: A peer-to-peer anonymizing network layer - Freedman, Morris - 2002 |

324 |
Random Geometric Graphs
- Penrose
- 2003
(Show Context)
Citation Context ... A d-dimensional RGG, denoted Gd(n, r), is obtained by placing n nodes uniformly at random on the surface of a d-dimensional unit torus, and connecting nodes within Euclidean distance r of each other =-=[34]-=-. RGGs have been studied in the context of random walks, and thus we can utilize some of these results for our purposes. Specifically, the G2(n, r) graph, also known as the Unit Disk graph, is often u... |

297 | Gossip-Based Ad Hoc Routing
- Haas, Halpern, et al.
- 2006
(Show Context)
Citation Context ...ted various approaches for disseminating data using several gossip func28 Te ch ni ons-sCo m pu te r S ci en cesD ep ar tm en t -sT ec hn ic alsR ep or tsCS -2 00 605s-s20 06 tions in ad hoc networks =-=[22]-=-. They investigate the impact of gossip on the message delivery ratio of broadcast messages. The anonymous gossip work has explored the use of gossip with direct neighbors in an ad hoc network in orde... |

281 | Geometric bounds for eigenvalues of Markov chains,” The Annals of Applied Probability - Diaconis, Stroock - 1991 |

268 | Random walks on graphs: a survey
- Lovász
- 1993
(Show Context)
Citation Context ...t is easy to check that π is also a stationary distribution of P , that is, πP = π, meaning that if the random walk happens to reach the distribution π, it stays there forever. A simple analysis (cf. =-=[28]-=-) shows that the stationary distribution of the simple random walk has a limit distribution that assigns probabilities to nodes proportionally to their degree: Proposition 3.1. The stationary limit di... |

251 | Lightweight probabilistic broadcast
- Eugster, Handurukande, et al.
- 2001
(Show Context)
Citation Context ... information. Instead, they only require each member to hold a partial random view of the network membership. Examples of such applications are probabilistic reliable dissemination of data and events =-=[5, 14, 27]-=-, peer sampling services [26], location services and uniform quorums [23], random overlay constructions [31], DHTs in which the hash function is calculated on node ids [35], etc. Therefore, it makes s... |

202 | Bimodal multicast
- Birman, Hayden, et al.
- 1999
(Show Context)
Citation Context ... information. Instead, they only require each member to hold a partial random view of the network membership. Examples of such applications are probabilistic reliable dissemination of data and events =-=[5, 14, 27]-=-, peer sampling services [26], location services and uniform quorums [23], random overlay constructions [31], DHTs in which the hash function is calculated on node ids [35], etc. Therefore, it makes s... |

187 | A.: Random walks in peer-to-peer networks
- Gkantsidis, Mihail, et al.
- 2004
(Show Context)
Citation Context ...small increases of the TTL may increase the search space exponentially. The work in [30] reported attractive empirical results, but does not provide any analytical evaluation of the RW properties. In =-=[18]-=-, Gkantsidis et al. explore the performance of random walks for searching and uniform sampling in peer-to-peer networks. An important result of this paper is that it is possible to simulate the select... |

187 | Symphony: Distributed hashing in a small world - Manku, Bawa, et al. - 2003 |

177 | Improved bounds for mixing rates of Markov chains and multicommodity flow - Sinclair - 1992 |

162 | CYCLON: Inexpensive Membership Management for Unstructured P2P Overlays - Voulgaris, Gavidia, et al. |

155 | The peer sampling service: Experimental evaluation of unstructured gossip-based implementations
- Jelasity, Guerraoui, et al.
- 2004
(Show Context)
Citation Context ...uire each member to hold a partial random view of the network membership. Examples of such applications are probabilistic reliable dissemination of data and events [5, 14, 27], peer sampling services =-=[26]-=-, location services and uniform quorums [23], random overlay constructions [31], DHTs in which the hash function is calculated on node ids [35], etc. Therefore, it makes sense to offer an optimized me... |

144 | Sequential analysis: tests and confidence intervals - Siegmund - 1985 |

143 | Probabilistic Reliable Dissemination in LargeScale Systems - Kermarrec, Massoulié, et al. |

110 | Anonymous gossip: Improving multicast reliability in mobile ad-hoc networks
- Chandra, Ramasubramanian, et al.
- 2001
(Show Context)
Citation Context ...io of broadcast messages. The anonymous gossip work has explored the use of gossip with direct neighbors in an ad hoc network in order to increase the reliability of broadcast and multicast protocols =-=[8]-=-. Both these works, however, do not address membership maintenance. 8 Discussion In this paper we have presented RaWMS, a random walk based lightweight membership service for ad hoc networks. We have ... |

106 | Route driven gossip: Probabilistic reliable multicast in ad hoc networks
- Luo, Eugster, et al.
- 2003
(Show Context)
Citation Context ...d the minimal and maximal view sizes of all nodes. A gossip-based membership service for sensor networks and for mobile ad hoc networks is described in [17]. We discuss it in detail in Section 5. RDG =-=[29]-=- is an adaptation of [14] to ad hoc networks. It reduces the cost of routing compared to [14] by utilizing routes created by other applications running in the same wireless node or by using proactive ... |

106 | Anonymous web transactions with crowds - Reiter, Rubin - 1999 |

98 | Scamp: Peer-to-peer lightweight membership service for large-scale group communication
- Ganesh, Kermarrec, et al.
- 2001
(Show Context)
Citation Context ...rvice. In particular, a general gossip-based peer sampling service was introduced in [26]. Examples of gossip-based lightweight membership services complying to this general framework are reported in =-=[1, 14, 16]-=-. We discuss gossip-based algorithms for membership construction in more detail in Section 5. SCAMP [16] introduced a generic random membership service that is used for probabilistic reliable dissemin... |

97 | Fastest mixing Markov chain on a graph
- Boyd, Diaconis, et al.
(Show Context)
Citation Context ...e random walk directly to obtain uniform sampling of network nodes. Instead, we use a different random walk, called the Maximum Degree (MD) random walk, which has been used before in various contexts =-=[28, 4, 6, 7]-=- to achieve uniform sampling. Let G = (V,E) be an undirected, connected, and non-bipartite graph, which is not necessarily regular. Suppose we have an upper bound D on dmax, the maximum degree of G.3 ... |

97 | Constrained Random Walks on Random Graphs: Routing Algorithms for Large Scale Wireless Sensor Networks”, WSNA
- Servetto, Barrenechea
- 2002
(Show Context)
Citation Context ... is required in order to determine the length of the RW in the reverse sampling procedure (the mixing time). There are a number of methods for obtaining a loose upper bound on the network size, e.g., =-=[15, 36]-=-. Once we have such a loose upper bound, we can use the birthday paradox principle [32] to obtain a much tighter bound. This is by keeping track of the average time between publications of the same id... |

93 | M.: Gossipbased peer sampling - Jelasity, Voulgaris, et al. - 2007 |

88 | T-man: Gossip-based overlay topology management - Jelasity, Babaoglu - 2005 |

77 | Approximating Aggregate Queries about Web Pages via Random Walks
- Bar-Yossef, Berg, et al.
(Show Context)
Citation Context ...e random walk directly to obtain uniform sampling of network nodes. Instead, we use a different random walk, called the Maximum Degree (MD) random walk, which has been used before in various contexts =-=[28, 4, 6, 7]-=- to achieve uniform sampling. Let G = (V,E) be an undirected, connected, and non-bipartite graph, which is not necessarily regular. Suppose we have an upper bound D on dmax, the maximum degree of G.3 ... |

68 | Random Walk for Self-Stabilizing Group Communication in Ad Hoc Networks
- Dolev, Schiller, et al.
- 2006
(Show Context)
Citation Context ...s on random geometric graphs have been investigated by [7] and [2]. These include computing the mixing time and the partial cover time. We rely on these results in our work. Dolev, Schiller and Welch =-=[12]-=- propose a randomized self-stabilizing group membership service for ad hoc networks. The group membership list is carried through the network by a random walk of a single mobile agent. The agent trave... |

68 | Random graphs, 2 nd ed - Bollobás - 2001 |

62 | Newscast computing
- Jelasity, Kowalczyk, et al.
- 2003
(Show Context)
Citation Context ... in [26] that lpbcast has a non-negligible chance of partitioning. Moreover, when partitions do occur in lpbcast, or any other similar gossip algorithm, they cannot self-heal. 5.2.2 Newscast Newscast =-=[25]-=- corresponds to (rand, pushpull, full, head). Newcast posses the same properties and drawbacks of lpbcast. Namely, it uses routing, it is not fully uniform, it suffers from partitioning, and there is ... |

55 | Ekta: An Efficient DHT Substrate for Distributed Applications
- Pucha, Das, et al.
- 2004
(Show Context)
Citation Context ...n of data and events [5, 14, 27], peer sampling services [26], location services and uniform quorums [23], random overlay constructions [31], DHTs in which the hash function is calculated on node ids =-=[35]-=-, etc. Therefore, it makes sense to offer an optimized membership service that indeed only provides nodes with partial random views. Such optimized services are the focus of this paper. 1 Te ch ni ons... |

51 | Large-scale newscast computing on the Internet - Jelasity, Steen - 2002 |

46 | Correctness of a gossip based membership protocol
- Allavena, Demers, et al.
- 2005
(Show Context)
Citation Context ...en the resulting network is asymptotically connected with probability one, if and only if c(n) → ∞ as n → ∞. In [33], the authors obtain a similar result when nodes are distributed in the unit square =-=[0, 1]-=-2. Throughout this paper we assume a radius r = √ C lnn n for the transmission range of nodes, where C is a constant. For C > 1/π, this radius satisfies the connectivity condition of Gupta and Kumar. ... |

45 | Peer counting and sampling in overlay networks : Random walk methods - Massoulie, Merrer, et al. - 2006 |

38 | Bluetooth and Wi-Fi Wireless Protocols: A survey and comparison - Ferro, Potorti - 2005 |

33 | Mixing times for random walks on geometric random graphs
- Boyd, Ghosh, et al.
(Show Context)
Citation Context ...e random walk directly to obtain uniform sampling of network nodes. Instead, we use a different random walk, called the Maximum Degree (MD) random walk, which has been used before in various contexts =-=[28, 4, 6, 7]-=- to achieve uniform sampling. Let G = (V,E) be an undirected, connected, and non-bipartite graph, which is not necessarily regular. Suppose we have an upper bound D on dmax, the maximum degree of G.3 ... |

31 | Matrix analysis. Cambridge university press - HORN, JOHNSON - 2012 |

29 | Ad hoc mobility management with randomized database groups
- Haas, Liang
- 1999
(Show Context)
Citation Context ...ew of the network membership. Examples of such applications are probabilistic reliable dissemination of data and events [5, 14, 27], peer sampling services [26], location services and uniform quorums =-=[23]-=-, random overlay constructions [31], DHTs in which the hash function is calculated on node ids [35], etc. Therefore, it makes sense to offer an optimized membership service that indeed only provides n... |

29 | Araneola: A Scalable Reliable Multicast System for Dynamic Environments
- Melamed, Keidar
- 2004
(Show Context)
Citation Context ...les of such applications are probabilistic reliable dissemination of data and events [5, 14, 27], peer sampling services [26], location services and uniform quorums [23], random overlay constructions =-=[31]-=-, DHTs in which the hash function is calculated on node ids [35], etc. Therefore, it makes sense to offer an optimized membership service that indeed only provides nodes with partial random views. Suc... |

21 |
On the transmission range in dense ad hoc radio networks
- PANCHAPAKESAN, MANJUNATH
(Show Context)
Citation Context ...k and each node transmits at a power level that covers an area of πr2 = log n+c(n)n , then the resulting network is asymptotically connected with probability one, if and only if c(n) → ∞ as n → ∞. In =-=[33]-=-, the authors obtain a similar result when nodes are distributed in the unit square [0, 1]2. Throughout this paper we assume a radius r = √ C lnn n for the transmission range of nodes, where C is a co... |

18 | On the connectivity of random m-orientable graphs and digraphs - Fenner, Frieze - 1982 |

15 | Rapidly mixing markov chains: A comparison of techniques. Available at http://www. cs.washington.edu/homes/venkat/pubs/pubs.html
- Guruswami
- 2000
(Show Context)
Citation Context ...n(P ) > −1. Hence, for a connected non-bipartite graph, 1 − λmax(P ) > 0. The difference 1− λmax(P ) is called the spectral gap of P and turns out to determine the mixing time of the random walk (cf. =-=[21]-=-): Theorem 3.3. The mixing time of a random walk with transition matrix P is upper bounded as follows: Tmix(ǫ) ≤ lnπ −1 min + ln ǫ −1 1− λmax(P ) , where πmin = min{π(v) | v ∈ V }. When π is the unifo... |