Roussi Roussev

Towards a self-managing software patching process using black-box persistent-state manifests

We describe an approach to self-managing software patching. We identify visibility into patch impact as the key missing component in automating the current patching process, and we present a suite of components that provides this visibility by constructing black-box persistent-state manifests through self-monitoring of dependencies. Additionally, we use the component suite to measure the actual impact of recent patches on several important commercial applications.

Analyzing persistent state interactions to improve state management

A primary challenge to building reliable and secure computer systems is managing the persistent state of the system: all the executable files, configuration settings and other data that govern how a system functions. The difficulty comes from the sheer volume of this persistent state, the frequency of changes to it, and the variety of workloads and requirements that require customization of persistent state. The cost of not managing a system‘s persistent state effectively is high: configuration errors are the leading cause of downtime at Internet services, troubleshooting configuration problems is a leading component of total cost of ownership in corporate environments, and malware— effectively, unwanted persistent state—is a serious privacy and security concern on personal computers. In this paper, we analyze how computer systems dynamically interact with files and configuration settings in an attempt to gain insights into the problem of persistent state management. We analyze over 3648 machine days of these persistent state interactions, collected over an 8 month period from 193 machines. These machines are under real workloads and include Internet servers, corporate desktops, and home machines. We characterize the scope and magnitude of the persistent state management problem today, measuring not only the gross characteristics of persistent state, but also analyzing how it is used by applications, and when administrators and users modify it. We find that monitoring persistent state interactions provides important visibility and show how it can be used as a foundation for building better persistent state management tools.