Input-sensitive performance testing

Abstract

One goal of performance testing is to find specific test input data for exposing performance bottlenecks and identify the methods responsible for these performance bottlenecks. A big and important challenges of performance testing is how to deeply understand the performance behaviors of a nontrivial software system in terms of test input data to properly select the specific test input values for finding the problematic methods. Thus, we propose this research program to automatically analyze performance behaviors in software and link these behaviors with test input data for selecting the specific ones that can expose performance bottlenecks. In addition, this research further examines the corresponding execution traces of selected inputs for targeting the problematic methods.