Near-optimal knowledge-free resilient leader election. (December 2022)

• Record Type: Journal Article
• Authors: Mo, Yuanqiu; Audrito, Giorgio; Dasgupta, Soura; Beal, Jacob
• Is Part Of: Automatica. Volume 146(2022)
• Publication Date: 2022-12
• Subjects: Leader election -- Multiagent system -- Resilience -- Aggregate computing -- Nonlinear feedback control -- Global stability; Automatic control -- Periodicals; Automation -- Periodicals; 629.805
• DOI: 10.1016/j.automatica.2022.110583
• Languages: English
• ISSNs: 0005-1098

Abstract

Abstract: Leader election, is a fundamental coordination problem in distributed systems. It has been addressed in many ways for different systems. Among these approaches, resilient leader election algorithms are of particular interest due to the ongoing emergence of open, complex distributed systems such as smart cities and the Internet of Things. However, previous algorithms attaining the optimal scaling of O(diameter) stabilization time complexity either assume some prior knowledge of the network or else that very large messages can be sent. In this paper, we present a resilient leader election algorithm with O(diameter) stabilization time, small messages, and no prior knowledge of the network. This algorithm is based on aggregate computing, which provides a layered approach to algorithm development based on composition of resilient algorithmic "building blocks." With our algorithm, a key design function g ( ⋅ ) defines important performance attributes: a fast-growing g ( ⋅ ) will delay discarding of obsolete data, while a slow-growing g ( ⋅ ) will slow down convergence to a single leader. We prove that the best asymptotic behavior for g ( x ) is ( 1 + 2 ) x + o ( x ), guaranteeing a near-optimal time complexity of ( 2 + 2 2 ) diameter + o(diameter) rounds for stabilization.