Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about Picsou, a novel system that enables efficient communication between replicated state machines through this 16-minute conference presentation from OSDI '25. Discover how researchers from UC Berkeley, University of Oregon, and University of Michigan address the fundamental problem that replicated state machines (RSMs) cannot communicate effectively due to the lack of formal frameworks and efficient protocols. Explore the introduction of Cross-Cluster Consistent Broadcast (C3B), a new primitive that forms the foundation of Picsou's practical implementation. Understand how Picsou draws inspiration from networking protocols like TCP to achieve constant metadata overhead in failure-free scenarios and minimize message resends during failures. Examine the system's flexibility in supporting both crash fault tolerant and Byzantine fault tolerant protocols for RSM communication. Delve into the core concept of Quacks (quorum acknowledgments) that enables nodes in each RSM to precisely determine message delivery status and identify likely message losses. Review performance evaluation results demonstrating up to 24× improvement over existing solutions across various applications including disaster recovery and data reconciliation scenarios.
