Project Grant 2552236

The National Science Foundation Division of Information and Intelligent Systems awarded the University of Pennsylvania $334,000 on August 1, 2026, under the Computer and Information Science and Engineering program (CFDA 47.070) to study metastable failures in large distributed computer systems.

The research develops a unified framework addressing self-sustaining failure cycles where system degradation persists even after the original triggering problem is resolved. The first thrust establishes formal models, definitions, and empirical foundations explaining how feedback loops among components create persistent degradation. The second thrust designs detection and mitigation primitives, including visualization techniques, reduced system representations, formal logic, and verification methods enabling rigorous reasoning about metastability. The third thrust focuses on scalability by extracting predictive models from logs and telemetry, identifying vulnerable states, generating triggering conditions, and validating the approach on realistic workloads. The project delivers open-source tools, datasets, and benchmarks alongside educational materials and training activities.

Work is performed in Philadelphia, Pennsylvania, with performance extending through July 31, 2029. The research strengthens reliability of critical digital infrastructure supporting cloud computing, online commerce, financial services, and social platforms by enabling earlier detection and prevention of prolonged outages and economic losses.