Modeling Distributed Applications for Mobile Ad Hoc Networks Using Attributed Task Graphs

Abstract

Abstract-Mobile ad hoc networks (MANETs) have received significant attention from the research community recently owing to the growth in popularity of portable computing and wireless networking. While researchers have primarily focused on developing lower layer mechanisms such as channel access and routing for making MANETs operational, higher layer issues such as application modeling have largely remained ignored. In this chapter, we present a novel distributed application framework based on attributed task graphs that enables a large class of resource discovery based applications on mobile, failure-prone environments such as MANETs. A distributed application is represented as a task comprised of smaller sub-tasks that need to be performed on different classes of computing devices with specialized roles. Execution of a particular task on a MANET requires several logical patterns of dataflow between nodes representing such device classes. As a result, dependencies are induced between the different classes of devices that need to cooperate to execute the application. Such dependencies yield a task graph representation of the distributed application.We consider the problem of executing distributed tasks on a MANET by means of dynamic selection of specific devices that are needed to execute the sub-tasks. We present simple and efficient algorithms for dynamic discovery and selection of suitable devices in a MANET from among a number of them providing the same functionality. This is carried out with respect to the proposed task graph representation of the application, and we call this process task embedding or anycasting. Since MANETs are prone to disconnections, we advocate periodic monitoring of the selected devices by one another. In the event of an application disruption owing to node mobility or failures, our algorithms adapt to the situation and dynamically rediscover the affected parts of the task graph, if possible. We propose metrics for evaluating the performance of these algorithms and report simulation results for a variety of application scenarios differing in complexity, traffic, and device mobility patterns. We demonstrate by simulation that our protocol can instantiate and reinstantiate TG nodes effectively in mobile scenarios; also the delivered effective throughput is near perfect at low to medium degrees of mobility and moderately high for high mobility scenarios.