IsoPredict: Dynamic Predictive Analysis for Detecting Unserializable Behaviors in Weakly Isolated Data Store Applications

Explore a 21-minute video presentation from the PLDI 2024 conference introducing IsoPredict, a novel dynamic predictive analysis tool for detecting unserializable behaviors in weakly isolated data store applications. Learn how researchers from Ohio State University developed this groundbreaking approach to identify potential errors in distributed data stores using weak isolation levels. Discover the innovative techniques employed by IsoPredict to handle divergent application behavior, solve mutually recursive constraint sets, and balance coverage, precision, and performance. Gain insights into the evaluation process that demonstrated IsoPredict's effectiveness in finding unserializable behaviors across four data store benchmarks, with over 99% of predicted executions proving feasible. Access supplementary materials, including reusable artifacts, to further understand the research methodology and findings.