WebPerf: Evaluating “What-If” Scenarios for Cloud-hosted Web Applications

Developers deploying web applications in the cloud often need to determine how changes such as service tiers or runtime loads may affect user-perceived page load time. We devise and evaluate a systematic methodology for exploring such “what-if” questions at the time a web application is deployed. Given a website, a web request, and “whatif” scenario, with a hypothetical configuration and runtime condition, our methodology, embedded in a system called WebPerf, can estimate a distribution of end-to-end response times for the request under the “what-if” scenario. WebPerf makes three contributions: (1) automated instrumentation of web sites written with increasingly popular task parallel libraries, to capture causal call dependencies of various computation and asynchronous I/O calls; (2) an algorithm to use the call dependencies, together with online- and offlineprofiled models of various I/O calls to estimate a distribution of end-to-end latency of the request; and (3) an algorithm to optimize modeling errors by deciding how many measurements to take within a limited time. We have implemented WebPerf for Microsoft Azure. Our experiments with five real websites and seven scenarios show that the median error of WebPerf’s estimation is within 7% for all applications and scenarios.