Collaborative Research: EAGER: Automating CI Configuration Troubleshooting with Bayesian Group Testing

在大规模网络基础设施（CI）软件系统中，由于存在大量可配置参数，配置故障诊断是一项复杂而昂贵的任务。日志挖掘和机器学习分析等现有方法在这种环境中面临挑战。为了解决这个问题，我们提出了BGT4AutoCI（使用贝叶斯组测试自动CI配置故障排除），这是一个利用贝叶斯组测试的突破性解决方案，即使在最小的先验知识和测试错误下也能确保准确的结果。经验丰富的CI操作员可以用近似的先验知识加快流程。本研究旨在彻底改变CI配置故障排除，引入一种高度精确和高效的方法，以优化当前和未来大规模CI系统的利用率。本研究的主要焦点是解决大规模CI系统中自动化配置故障排除中的关键挑战。提出的三层方法包括：(1)为CI配置故障排除制定贝叶斯组测试，该方法采用格模型在单个配置参数级别准确识别风险，并在故障排除过程中考虑不确定性。(2)贝叶斯群测试的多项范式，它引入多项响应来同时观察CI系统的多个方面，从而实现有效的测试选择算法，以联合测试影响CI各个方面的配置参数。(3)自动配置故障排除，涉及几个关键组件的设计，以建立BGT4AutoCI作为一个自动配置故障排除框架，最大限度地减少对人工干预的需求。本项目的成果有可能显著提高CI配置故障排除的效率和准确性，有利于当前和未来的大规模CI系统。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Configuration troubleshooting in large-scale cyberinfrastructure (CI) software systems is a complex and costly task due to numerous configurable parameters. Existing methods like log mining and machine learning analysis face challenges in such environments. To address this, we present BGT4AutoCI (Automating CI Configuration Troubleshooting with Bayesian Group Testing), a groundbreaking solution that leverages Bayesian Group Testing, ensuring accurate results even with minimal prior knowledge and testing errors. Experienced CI operators can expedite the process with approximated prior knowledge. This research aims to revolutionize CI configuration troubleshooting, introducing a highly precise and efficient approach that will optimize the utilization of current and future large-scale CI systems.The primary focus of this research is to address critical challenges in automated configuration troubleshooting within large-scale CI systems. The proposed three-fold approach encompasses: (1) Formulating Bayesian Group Testing for CI configuration troubleshooting, which employs lattice models to accurately identify risks at the individual configuration parameter level, taking uncertainty into account during troubleshooting. (2) A multinomial paradigm for Bayesian Group Testing, which introduces multinomial responses to simultaneously observe multiple aspects of CI systems, enabling efficient test selection algorithms for jointly testing configuration parameters that impact various aspects of CIs. (3) Automated configuration troubleshooting, which involves the designs of several key components to establish BGT4AutoCI as an automated configuration troubleshooting framework that minimizes the need for human intervention. The outcomes of this project hold the potential to significantly enhance the efficiency and accuracy of CI configuration troubleshooting, benefiting current and future large-scale CI systems.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.