D. The Problem Statement
Our goal is to automate finding performance problems by
executing the AUT on a small set of randomly chosen test
input data, and then inferring rules with a high precision
for selecting test input data automatically to find more
performance problems in the AUT. Specifically, these are
if-then rules that describe properties of input data that
result in good performance test cases that lead to increased
computational workload on applications when compared to
bad performance test cases when computational workload
is much smaller. For example, a rule may say “if inputs
convictedFraud is true and deadboltInstalled
is false then the test case is good.” In this work, we supply
automatically learned rules using a feedback mechanism
to test scripts, and these scripts parse these rules and use
them to guide test input data selection automatically to
steer execution of the AUT towards code that has exposes
performance problems.