On Design, Evaluation and Enhancement of IP-Based Routing Solutions for Low Power and Lossy Networks

In early 2008, a new IETF Working Group (WG), namely ROLL, was chartered to investigate the suitability of existing IP routing protocols for Low Power Lossy Networks (LLNs), which at the time were suffering compatibility issues due to the pervasive use of proprietary protocols. Given the vision of the Internet of Things (IoT) and the role LLNs would play in the future Internet, the IETF set out to standardize an IPv6 based routing solution for such networks. After surveying existing protocols and determining their unsuitability, the WG started designing a new distance vector protocol called RPL (recently standardized in IETF RFC 6550) to fulfill their charter. Joining the WG efforts, we developed a very detailed RPL simulator and using link and traffic traces for existing networks, contributed with a performance study of the protocol with respect to several metrics of interest, such as path quality, end-to-end delay, control plane overhead, ability to cope with instability, etc. This work was standardized as IETF Informational RFC 6687.This detailed study uncovered performance issues for networks of very large scale. In this thesis, we provide an overview of RPL, summarize our findings from the performance study, analysis and comparison with a reactive lightweight protocol and suggest modifications to the protocol that yield significant performance improvements with respect to control overhead and memory consumption in very large scale networks. For future work, we propose a routing technique, named Hybrid Intelligent Path Computation (HIPC), along with modifications to the original RPL protocol standard, that outperforms solely distributed or centralized routing techniques. Finally, we also show how one can facilitate Quality of Service (QoS), load balancing and traffic engineering provision in the IoT without incurring any extra control overhead in number of packets other than that already consumed by the proposed IETF standard, using a combination of centralized and distributed computation.Ph.D., Computer Science -- Drexel University, 201