Reliable Bulk Data Dissemination in Sensor Networks

RELIABLE BULK DATA DISSEMINATION IN SENSOR NETWORKS Leijun Huang, PhD George Mason University, 2007 Dissertation Director: Dr. Sanjeev Setia A wireless sensor network consists of a large number of resource-constrained sensor nodes that are self-organized into a multi-hop network and cooperate on a single task. In many situations, sensor networks need to run for a long time once deployed. When the environment changes during their lifetime, updating the code image or application data at the node for a new task becomes necessary, thus making data dissemination a critical issue where a large data object needs to be reliably propagated to all of the nodes in a network. While most of the current sensor nodes are equipped with a multiple-channel radio, the existing data dissemination approaches such as Deluge [1] do not take advantage of multiple channels. Moreover, these approaches mostly focus on the object delivery latency, while energy efficiency is also very important due to the resource constraints of the sensor nodes. This dissertation proposes three novel protocols for reliable bulk data dissemination, named McTorrent, CORD and McCORD, that focus on both object delivery latency and energy efficiency. These protocols use multiple channels, or a core-based two-phase approach, or both techniques to reduce object delivery latency and energy consumption at each node. The results from experiments on both indoor and outdoor testbeds and extensive simulations in various scenarios show that these protocols significantly reduce the latency and/or energy consumption, compared to the existing approaches. Chapter 1: Introduction A wireless sensor network consists of a large number of resource-constrained sensor nodes that are self-organized into a multi-hop network and cooperate on a single task. Examples of sensor network applications include environmental monitoring, perimeter surveillance and product inventory maintenance. In many situations, sensor networks need to run for a long time once deployed. When the environment changes during their lifetime, updating their code image (the software that they are running) for a new task becomes necessary. Due to the large number of the deployed sensor nodes and the environment where physically collecting previously deployed nodes is either very difficult or infeasible, the code image needs to be updated “over-the-air”, using the existing network itself. In addition to the code image, application data may also require updating. For example, in a target detection application, the sensor nodes use the image of the target stored in their memory to detect the target. When the target changes, the image of the new target needs to be propagated to the network. The critical service required for these purposes is a reliable bulk data dissemination protocol, i.e., a protocol for reliably distributing a large object to all the nodes in the network. The reliable bulk data dissemination problem is a special case of the reliable downstream data transport problem for sensor networks [2–4]. The main challenge in bulk data dissemination is that the size of the object being distributed is two to three orders of magnitude larger than the size of a data packet, and is also much larger than the amount of RAM