Fundamental control algorithms in mobile networks

Abstract

In this work we propose simple and efficient protocols for counting and leader election in mobile networks. For mobile networks with fixed base stations we provide a new and very efficient protocol for counting the number of mobile hosts. The main part of the work concentrates on ad-hoc networks (no fixed subnetwork).We provide a model for these networks and leader election (and a special form of counting) protocols for both named and anonymous mobile hosts. In this work we define two protocol classes, the Non-Compulsoryprotocols, which do not affect the motion of the hosts and the Compulsory, which determine the motion of some or all the hosts. By assuming that the mobile hosts move as if each one is doing a continuous random walk on their allowable space S of motions, and by assuming a universal time, we show that our leader election protocol terminates (with high probability and also on the average) in time asymptotically linear to the size of the space S, measured as its volume divided by the volume of the sphere defined by the range of transmission of each mobile host. We also provide a simple but very efficient Compulsory (forced random walks) Las Vegas protocol for leader election in ad-hoc networks, which also allows counting, with termination detection.Our analysis techniques for the meeting time of concurrent random walks extend the known facts and are tight. They may be used as an analysis tool in the design of many other distributed protocols. This is the first algorithmic and characterization work, to our knowledge, for ad-hoc networks.

A Stable CoveringSet-based Leader Election Algorithm (SCLEA) for Mobile Ad hoc Distributed Systems

“…In [18], an extrema-finding leader election algorithm for mobile ad hoc networks had been proposed. Although designing an algorithm with an extrema-finding aspect is interesting for environments such as ad hoc networks (because it is important to select a leader with performance-related attributes as computational capabilities, battery power or nodes' velocity), the algorithms in [18] are considered to be unrealistic and not suitable for some applications since all nodes (without any exception) are required to exchange information in order to elect a leader, and this is impractical.…”

“…The approach of [15] is exposed to frequent link breakages and failures of nodes that can occur during or after election, while the two protocols proposed in [16] have been provided with no proofs of correctness. The approaches provided in [17] and [18] are both impractical since the former assumes that all nodes that participate in election process have the same evaluation function, while the later assumes that all nodes (without any exception) are required to exchange information in order to elect a leader, and this is not realistic!…”

A leader election algorithm for dynamic networks with causal clocks

“…Here we mention some representative papers on leader election in dynamic networks. Hatzis et al [12] presented algorithms for leader election in mobile networks in which nodes are expected to control their movement in order to facilitate communication. This type of algorithm is not suitable for networks in which nodes can move arbitrarily.…”

Design of an Leader Election Protocol in Mobile Ad Hoc Distributed Systems

“…Leader election algorithms for mobile ad hoc networks have been proposed in [9] [10]. As noted earlier, we are interested in an extrema-finding algorithm, because it is desirable to elect a leader with some system-related attributes such as maximum battery life or maximum computation power.…”

“…The algorithms in [9] are not extrema-finding and cannot be extended to perform extrema finding. Although, extrema-finding leader election algorithms for mobile ad hoc networks have been proposed in [10], these algorithms are unrealistic as they require nodes to meet and exchange information in order to elect a leader and are not well-suited to the applications discussed earlier. Several clustering algorithms have been proposed for mobile networks [11][12], but these algorithms elect cluster-heads only within their single hop neighborhood.…”