CRII: CSR: Toward Understanding and Automatically Detecting Specious Configuration in Large Systems

Johns Hopkins University, Baltimore MD

Investigators

Abstract

Software configurations allow users to customize a software system for different needs. Today, it is not uncommon for a system to expose hundreds of parameters to users for customizing configuration. Such high customizability comes at a price, as errors in configuration are a major source of system failures. While various solutions have been developed to tackle misconfiguration, they mainly focus on invalid configurations, which violate some correctness criteria. Unfortunately, many misconfigurations are technically valid, yet cause severe production problems such as poor performance. These are called specious configurations. The goal of this project is to systematically investigate and proactively prevent specious configurations. Specifically, the project will study specious configuration from both open-source and commercial systems to understand its broad scope, traits and how it affects system execution. Informed by this study, this project will then design algorithms and techniques for modeling the risk of a configuration change by using static analysis and selective symbolic execution to construct parameterized impact tables and then enhance them with offline profiling data. Based on this model, this project will develop an online risk assessment tool which can alert expert operators about specious configuration before production. Specious configuration is frequently cited for lost productivity. This project will further our understanding of underlying factors behind specious configuration and help practitioners reduce misconfiguration-induced failures in large-scale systems. Results from this research will be distilled into graduate seminars and classes on Operating Systems. The resulting dataset, software, publications and course materials will be maintained in an open-source repository for at least five years beyond the completion of the project. 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.