Towards Understanding Systems Through User Interactions

Modern computer systems are complex. Even in the best of conditions, it can be difficult to understand the behavior of the system and identify why certain actions are occurring. Existing systems attempt to provide insight by reviewing the effects of actions on the system and estimating their cause. As computer systems are strongly driven by actions of the user, we propose an approach to identify processes which have interacted with the user and provide data to which system behaviors were caused by the user. We implement three sensors within the graphical user interface capable of extracting the necessary information to identify these processes. We show our instrumentation is effective in characterizing applications with an on-screen presence, and provide data towards the determination of user intentions. We prove that our method for obtaining the information from the user interface can be done in an efficient manner with minimal overheads