CAREER: Rethinking Configuration Management for Cloud and Datacenter Systems

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

Configuration management is an integral part of modern cloud/datacenter system management. Many critical operations update production system configurations (e.g., performance tuning and feature rollout). Today, large-scale cloud and Internet services evolve rapidly, with hundreds to thousands of configuration changes daily. With the high velocity of changes, faulty configuration changes inevitably become major causes of system failures and service outages, leading to catastrophic impacts. This project aims to (1) develop novel, practical technologies to address key problems of existing techniques, and (2) explore radically new system designs to prevent configuration problems. The project will develop scientific foundations, new methodologies, techniques, and practical tools for cloud system configuration management, based on four synergistic research thrusts. First, it will establish the foundation of configuration testing, a key missing piece of configuration management today. Second, it will develop novel, practical configuration testing techniques to prevent failure-inducing configuration changes. Third, it will address the key problems of static configuration validation, the de facto pre-deployment protection. Fourth, it will explore new configuration designs that enable systems effectively to repair and recover from faulty configurations and new interfaces to advance configuration manageability significantly and reduce misconfigurations. The project will be a catalyst for multidisciplinary research and education on cloud computing, computer systems, and software engineering at the University of Illinois. It will expand course offerings by designing new interdisciplinary courses and by enhancing existing courses. The project is committed to broadening participation in computing and will engage undergraduate and under-represented students in research. The project will have major impacts on industry. The configuration technologies in practice are often insufficient, partially due to a lack of research foundation and rigorous, systematic techniques. The project’s industry collaboration will help drive practical solutions and transition them into practice. The project will be a strong advocate for open science. It will make the research artifacts openly accessible, reproducible, and reliable, including software prototypes, datasets, evaluation artifacts and scripts, and research results. All research artifacts will be actively maintained throughout the project period and will be accessible after an embargo period of one year past the conclusion of the project and any extensions. The project will have an index page (illiconfig.github.io) to help navigate all the research artifacts and the repositories will be backed up continuously using the Engineering Information Technology service at the University of Illinois. 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.