Generalized Wing-Gong algorithm finds bugs 370x faster in distributed storage

A team of researchers led by Kotikala Raghav has generalized the classic Wing-Gong (WG) linearizability checking algorithm to make consistency checking for distributed storage systems far more versatile and significantly faster. The technique, detailed in an arXiv paper (2608.17388), goes beyond standard linearizability checks by supporting other non-transactional consistency guarantees, including ordered sequential consistency as provided by Zookeeper. It also allows engineers to specify system-specific ordering constraints, so checkers can validate behavior against the exact semantics of their distributed system.

The researchers applied the generalized checker across 8 real distributed storage systems and measured dramatic improvements. Checking for system-specific consistency guarantees proved easy to implement, reduced false negatives during testing, and delivered up to 370x speedups over existing tools. This speedup also enables checking with more concurrent clients within the same time budget, making it practical to test workloads that were previously too slow to verify.

Most notably, the checker uncovered 6 new consistency violation bugs across the tested systems, and 5 of those could not be found by existing consistency checkers. This shows that the extended algorithm not only improves performance but also catches real-world bugs that prior linearizability-focused tools would miss entirely. By generalizing the WG approach, the work gives distributed systems engineers a more powerful, system-aware testing tool for ensuring correctness under concurrency.