Distributed consensus in wireless network

Yu, Dachao (2023) Distributed consensus in wireless network. PhD thesis, University of Glasgow.

Abstract

Connected autonomous systems, which are powered by the synergistic integration of the Internet of Things (IoT), Artificial Intelligence (AI), and 5G technologies, predominantly rely on a central node for making mission-critical decisions. This reliance poses a significant challenge that the condition and capability of the central node largely determine the reliability and effectiveness of decision-making. Maintaining such a centralized system, especially in large-scale wireless networks, can be prohibitively expensive and encounters scalability challenges. In light of these limitations, there’s a compelling need for innovative methods to address the increasing demands of reliability and latency, especially in mission-critical networks where cooperative decision-making is paramount. One promising avenue lies in the distributed consensus protocol, a mechanism intrinsic to distributed computing systems. These protocols offer enhanced robustness, ensuring continued functionality and responsiveness in decision-making even in the face of potential node or communication failures.

This thesis pivots on the idea of leveraging distributed consensus to bolster the reliability of mission-critical decision-making within wireless networks, which delves deep into the performance characteristics of wireless distributed consensus, analyzing and subsequently optimizing its attributes, specifically focusing on reliability and latency. The research begins with a fundamental model of consensus reliability in an crash fault tolerance protocol Raft. A novel metric termed ReliabilityGain is introduced to analyze the performance of distributed consensus in wireless network. This innovative concept elucidates the linear correlation between the reliability inherent to consensus-driven decision-making and the reliability of communication link transmission. An intriguing discovery made in my study is the inherent trade-off between the time latency of achieving consensus and its reliability. These two variables appear to be in contradiction, which brings further performance optimization issues.

The performance of the Crash and Byzantine fault tolerance protocol is scrutinized and they are compared with original centralized consensus. This exploration becomes particularly pertinent when communication failures occur in wireless distributed consensus. The analytical results are juxtaposed with performance metrics derived from a centralized consensus mechanism. This comparative analysis illuminates the relative merits and demerits of these consensus strategies, evaluated from the dual perspectives of comprehensive consensus reliability and communication latency.

In light of the insights gained from the detailed analysis of the Raft and Hotstuff BFT protocols, my thesis further ventures into the realm of optimization strategies for wireless distributed consensus. A central facet of this exploration is the introduction of a tailored communication resource allocation scheme. This scheme, rooted in maximizing the performance of consensus mechanisms, dynamically assesses the network conditions and allocates communication resources such as transmit power and bandwidth to ensure efficient and timely decision-making, which ensures that even in varied and unpredictable network conditions, consensus can be achieved with minimized latency and maximized reliability.

The research introduces an adaptive protocol of distributed consensus in wireless network. This proposed adaptive protocol’s strength lies in its ability to autonomously construct consensus-enabled network even if node failures or communication disruptions occur, which ensures that the network’s decision-making process remains uninterrupted and efficient, irrespective of external challenges. The sharding mechanism, which is regarded as an effective solution to scalability issues in distributed system, does not only aid in managing vast networks more efficiently but also ensure that any disruption in one shard cannot compromise the functionality of the entire network. Therefore, this thesis shows the reliability and security analysis of sharding that implemented in wireless distributed system. In essence, these intertwined strategies, rooted in the intricate dance of communication resource allocation, adaptability, and sharding, together form the bedrock of my contributions to enhancing the performance of wireless distributed consensus.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: T Technology > T Technology (General); T Technology > TA Engineering (General). Civil engineering (General)
• Colleges/Schools: College of Science and Engineering > School of Engineering
• Supervisor's Name: Zhang, Professor Lei and Imran, Professor Muhammad Ali
• Date of Award: 2023
• Depositing User: Theses Team
• Unique ID: glathesis:2023-83981
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 07 Dec 2023 09:27
• Last Modified: 07 Dec 2023 09:27
• Thesis DOI: 10.5525/gla.thesis.83981
• URI: https://theses.gla.ac.uk/id/eprint/83981

Related URLs

• Article DOI
• Article DOI
• Article DOI
• Article DOI