In this project we have studied a new type of algorithm that can outperform the current SOTA of Hot-Stuff and Kauri, called Pandemic Consensus Algorithm.
To execute the code we recommend to read the report-pdf and follow the instructions.
The expansion of our protocol follows a binomial tree structure, resembling a "pandemic" spread. Nodes act as small star leaders, broadcasting data to neighbors. Each node follows a formula to determine its broadcast target. Collision scenarios and data transmission reliability are addressed through mathematical induction.
Two communication methods are proposed: direct connection to the leader and data transmission via designated parent nodes, akin to a hierarchical dissemination.
A recovery system handles missing signatures by broadcasting messages to missing participants after a timeout. Future exploration includes optimizing timeouts and implementing finer-grained recovery mechanisms.
Four algorithms are implemented: Hot-Stuff, Kauri, and two variations of Binomial Tree. Each algorithm handles data transmission and consensus differently, with distinct recovery mechanisms.
Experiments compare protocol performance under various conditions. Metrics include latency, consensus achieved, and throughput. Results show Binomial Tree outperforms Kauri, particularly in speed and scalability.