Scalable, Self-Healing, and Real-Time Network Services for Directed Diffusion
Type of DegreeDissertation
DepartmentComputer Science and Software Engineering
MetadataShow full item record
Directed diffusion is a data-centric publish-subscribe routing protocol for sensor networks. We have proposed three network services which increase the capabilities of directed diffusion. Our protocols build on the inherent strengths of diffusion and lessen its weaknesses. The system architecture emphasizes efficient communication, local route repair, and real-time response. Our suite of network services significantly improves the performance of directed diffusion by addressing the fundamental challenges of sensor networks: energy-efficiency, dynamic environments, and scalability. Our design increases the efficiency of flooding, improves packet delivery rates in the presence of node failure, and decreases the number of packets that miss their deadlines. We evaluate the performance of our improved diffusion in terms of routing overhead, delivery effectiveness, and deadline achievement. Our results demonstrate the benefits of the network services in all three respects. We increase the efficiency of flooding by 48%, improve packet delivery rates in the presence of node failure by up to 28%, and decrease the number of packets that miss their deadlines by 30-60%.