Automatically Verifying Replication-Aware Linearizability

Explore a 14-minute conference presentation from OOPSLA 2025 that introduces a groundbreaking automated verification technique for replication-aware linearizability in distributed systems. Learn how researchers from IIT Madras, Microsoft Research, and Tarides have developed a novel approach to verify Mergeable Replicated Data Types (MRDTs) and state-based Conflict-free Replicated Data Types (CRDTs) that goes beyond traditional correctness specifications. Discover the limitations of strong eventual consistency as a correctness specification for Replicated Data Types and understand how replication-aware linearizability provides a more expressive alternative that relates converged states to updates received at replicas. Examine the novel algebraic properties identified for MRDT operations and merge functions that are sufficient for proving linearizability, extending beyond standard concepts of commutativity, associativity, and idempotence. Delve into the innovative bottom-up linearization technique that enables automatic verification of the required algebraic properties for both MRDTs and state-based CRDTs. See how this fully automated approach has been successfully applied to complex implementations, including a novel JSON MRDT, demonstrating its practical applicability in real-world distributed systems. Gain insights into the theoretical foundations of data replication for fault tolerance and low-latency access in decentralized applications, while understanding how this research advances the field of automated verification for distributed data structures. The presentation includes access to supplementary materials with artifacts that have been evaluated as available, reusable, and with reproduced results, providing comprehensive resources for further exploration of the verification techniques presented.