An Introduction to Visualizing Program Execution Dynamics with Chitra

We describe a tool called Chitra, which serves two purposes. First, Chitra visualizes program execution sequences (PES's) collected by monitoring parallel and distributed program execution in the time, event, and frequency domains. Second, Chitra produces an empirical model of a PES. A case study applies Chitra to predict and evaluate the efficiency and fairness of a resource sharing algorithm at a range of parameter values, given observations of a few parameter values. Inefficient behavior is explained by examining the states that constitute the stochastic process model constructed by Chitra.