CNS Core: Small: Automated testing for data- and compute-intensive distributed systems through feedback-based fuzzing

点击翻译按钮获取中文摘要

The demand for large-scale data storage and processing has made distributed systems crucial. However, programming modern distributed systems is particularly challenging because it requires correct reasoning about concurrent code running across machines, unreliable networks, software and hardware failures, and complex application logic. In practice, despite the challenges faced by developers, there are currently no good options to test distributed systems automatically and comprehensively. This project focuses on developing techniques that effectively test data- and compute-intensive distributed applications used by millions of users, running on virtually every cloud service and data center.By bringing together expertise in operating systems, distributed systems, and systems testing, this project will make contributions to both theory and practice by (1) designing and building efficient test execution engines that transparently make large distributed systems amenable to low-overhead and high-throughput execution, (2) exploring test coverage and mutation policies that are tuned to address multi-node and fault-tolerant systems, and (3) developing techniques that use the likely root cause of distributed system bugs to minimize test cases and reproduce production failures.Distributed systems are now the foundation of society's most essential services, including e-commerce, banking, financial, medical, and logistics services. This project will allow distributed system developers to build more reliable distributed systems by developing practical and high-coverage testing approaches. The testing approaches developed will prevent service downtime, data loss, system malfunctions, and business and customer trust loss across a range of multi-billion dollar industries. Furthermore, the testing systems developed will reduce developer effort and time devoted to fixing distributed systems, which are notoriously challenging for developers.All project data is stored in public sites and university storage systems to ensure safe long-term storage for at least seven years from the award conclusion or public release, whichever comes later. The data produced includes system implementations and source code, documentation, test sets, analysis datasets, and educational material. The project information will be located at https://www.cs.purdue.edu/homes/pfonseca/projects/ds-fuzzing.html.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.