Revisiting Asynchronous Rumor Spreading in the Blockchain Era

Asynchronous rumor spreading, or epidemic algorithms, are a class of data dissemination protocols that have been used throughout the years for a large variety of distributed applications. The emergence of large-scale, public blockchains, such as Bitcoin and Ethereum, has reinvigorated research interest in these protocols as they are employed to disseminate pending transactions and confirmed blocks in their peer-to-peer (P2P) networks. Efficient, timely and fault-tolerant information dissemination is vital for blockchain networks as it affects issues ranging from security to block finality. Recent works have analyzed the structural properties of blockchain network overlay graphs. Their findings show that they have inherent similarities to those of social networks, such as power-law degree distributions, small diameters and star-like communities. In this work, we present an experimental analysis of the vanilla asynchronous push & pull rumor spreading protocol that is employed by public blockchains. This protocol, although robust and scalable, can be substantially improved. We demonstrate this by analyzing the effect that multiple parameters have on the protocol's performance, such as using memory to avoid contacting the same neighbor twice in a row, varying the stopping criteria of nodes to decide when to stop spreading the rumor, employing more sophisticated neighbor selection policies instead of the standard uniform random choice and others. Prior works have focused on either providing theoretical upper bounds on the number of rounds needed to spread the rumor to all nodes, or, propose improvements by adjusting isolated parameters. To our knowledge, our work is the first to study how multiple parameters affect the protocol's behavior both in isolation and combination and under a wide range of values. Moreover, prior theoretical works have studied rumor spreading only on bidirectional social topologies. Our study examines the behavior of the protocol in multiple topology classes. These include bidirectional, directed, and also a special type of social topologies, called signed topologies, which resemble more closely the topologies of blockchain P2P networks. Our work is the first to indicate and deal with how chains of communities that are sparsely connected to the core of the network can hamper the rumor's spreading. Thus, we complement prior theoretical work to shed light on how the protocol behaves in practical, real-world, large scale distributed systems. Finally, through our detailed analysis, we demonstrate how a few simple additions to the protocol deliver a percentage decrease of the time required to inform all nodes by a maximum of 99.69% and an average of 86.04%.