Topiary: Fast, Scalable Publish/Subscribe for Peer-to-Peer (D)Apps

Abstract

The emergence of blockchain technology has fostered the development of numerous decentralized applications (dapps) in recent years Pub/sub (publish/subscribe) systems play a crucial role by associating messages with specific topics and propagating them from publishers to subscribers across the network. Decentralized pub/sub aims to provide this functionality without relying on centralized control or global network state information, enabling message propagation among nodes in a coordinated manner. Efficiency in pub/sub services entails ensuring that subscribers receive published messages promptly. We introduce Topiary, a rapid and scalable protocol designed for decentralized applications' pub/sub systems. Topiary autonomously learns an efficient peer-to-peer (p2p) topology tailored to the publish/subscribe network. It does so by analyzing peers' interactions with their neighbors. Inspired by concepts from the multi-armed bandit problem, Topiary strikes an optimal balance between maintaining connections with well-connected neighbors and exploring new connections within the network, based on their topical needs. Through experimental evaluations, Topiary has shown a 50% reduction in broadcast latency while achieving an interested topic coverage of over 98%, marking it as a promising solution for efficient decentralized pub/sub networks.