Improving performance and incentives in disruption-tolerant networks

Abstract

The recent proliferation of personal wireless devices has led to the emergence of disruption-tolerant networks (DTNs), which are characterized by intermittent connectivity among some or all participating nodes and a consequent lack of contemporaneous end-to-end paths between the source and consumer of information. However, the success of DTNs as a communication paradigm is critically dependent on the following challenges being addressed: (1) How to enable popular but demanding applications, such as video-on-demand, to operate in such constrained network settings, and (2) How to incentivize individual devices to cooperate when network operation is only possible under, or greatly benefits from cooperation.

In this dissertation, we present a novel set of protocols and develop real systems that effectively meet the above challenges. We make the following contributions:

First, we design and implement a novel system for enabling high bandwidth content distribution in vehicular DTNs by leveraging infrastructure access points (APs). We predict which APs will soon be visited by a vehicular node and then proactively push content-of-interest to those APs. Our replication schemes optimize content delivery by exploiting Internet connectivity, local wireless connectivity, node relay connectivity and mesh connectivity among APs. We demonstrate the effectiveness of our system through trace-driven simulation and Emulab emulation using real taxi and bus traces. We further deploy our system in two vehicular networks: a fourteen AP 802.11b network and a four AP 802.11n network with smartphones and laptops as clients.

Second, we propose an incentive-aware routing protocol for DTNs. In DTNs, routing takes place in a store-and-forward fashion with the help of relay nodes. If the nodes in a DTN are controlled by rational entities, such as people or organizations, the nodes can be expected to behave selfishly by attempting to maximize their utilities and conserve their resources. Since routing is inherently a cooperative activity, system operation will be critically impaired unless cooperation is incentivized. We propose the use of pair-wise tit-for-tat (TFT) as a simple, robust and practical incentive mechanism for DTNs. We then develop an incentive-aware routing protocol that allows selfish nodes to maximize their own performance while conforming to TFT constraints.