Comparative analysis of disruption tolerant network routing simulations in the ONE and ns-3

Abstract

This thesis studies the performance of Disruption Tolerant Networking (DTN) routing protocols, and the effect of simulator selection. Research into the Geo-location Assisted Predictive Routing (GAPR), and GAPR2 protocols at NPS used the ONE Simulator. The ONE abstracts everything below the routing layer to simplify the development of DTN protocols. In contrast, Network Simulator 3 (ns-3) simulates the entire network stack. ns-3 includes packet headers and existing link-layer protocols that the ONE abstracts away. The inclusion of link-layer overhead and packet headers reduces message delivery by 31% and increases average latency by 119%. Packets used to share routing information consume up to 33% of all transmitted data. Effective throughput between connected nodes decreases by 40%-70% of the equivalent ONE bandwidth. These penalties vary significantly depending on routing protocol design choices. This thesis implements Epidemic, Vector, Centroid, GAPR, and GAPR2 protocols in ns-3. It also combines Centroid with GAPR to create a new protocol called GAPR2a. The protocols are extensively simulated in three mobility scenarios in ns-3 and the ONE: one urban scenario and two military scenarios. GAPR2a provides the best overall performance in the urban scenario, and Vector provides the best overall performance in the military scenarios. Future DTN protocol development should continue in ns-3 because the ONE's abstractions may not reflect real-world performance.