CAREER: Resource Management in Mobile ad hoc Networks - the Spatial Dimension

University Of Texas At Dallas, Richardson TX

Investigators

Abstract

The goal of the proposed research is to develop a comprehensive understanding of the spatial aspect of distributed computing in mobile systems. Traditionally, events in a distributed computation are ordered by time. Scalar and vector clocks are used to represent logical time and causal dependencies. However, the mobility of nodes injects a new spatial dimension into the problem. The physical and network distances between nodes in a mobile network are not correlated. Moreover, some wireless links in a mobile network may be unidirectional. Hence, unlike physical distance, the network distance between nodes is not a symmetric relation. Finally, if requests of two mobile nodes are being prioritized on the basis of their respective distances from a third (resource) node, their relative priority can change with time. This is in contrast to the immutable relative ordering of events based on time. Hence, if distance is one of the criteria for ordering requests, existing solutions for problems like mutual exclusion (using logical clocks) cannot be efficiently used for resource allocation in mobile computing. We will develop a new formal representation of the spatial-temporal relationship between events in mobile computing. Space-time vectors are a step in this direction. A theory for a possibly immutable ordering of events based on their spatial-temporal coordinates will be developed. Such an ordering of requests could be used for resource deadlock free, starvation free and fair resource management. New protocols will be developed at the MAC sub-layer enabling nodes to communicate using unidirectional wireless links. We will develop new routing protocols that utilize this MAC sub-layer protocol, and can correctly route packets in a wireless mobile network with a combination of unidirectional and bidirectional links. Subsequently, the performance of these solutions will be evaluated through simulation experiments and prototyping. Graduate and undergraduate students will actively participate in every stage of the project.