PhD Defense: Finding Indicators of Load on a Simple Web Server via Analysis of Black-Box Performance Measurements
Traditional methods for system performance analysis have long relied on a mix of queuing theory, detailed system knowledge, intuition, and trial-by-error. These approaches often require construction of incomplete gray-box models that can be costly to build and difficult to scale or generalize. In this thesis, we present a robust method for black-box analysis of web server system performance with a focus on discovering the amount of load on a server with minimal knowledge of its internal mechanisms. In contrast to white-box analysis, where a system's internal mechanisms can help directly explain its behavior, black-box analysis relies on external measurements of a system's reaction to well-understood inputs. The primary advantages of black-box analysis are its relative independence from specific architectures, its applicability to opaque environments (e.g., closed-source systems), and its scalability. This work shows that statistical analyses of a simple external performance metric like response time can be used to discover which server resources are stressed by particular workloads, and whether the server is becoming saturated. We believe our approach presents the first steps toward a performance analysis regime that is useful for new paradigms like cloud computing.