System Performance Optimization via Design and Configuration Space Exploration

The runtime performance of a software system often depends on a large number of static parameters, which usually interact in complex ways to carry out system functionality and influence system performance. It's hard to understand such configuration spaces and find good combinations of parameter values to gain available levels of performance. Engineers in practice often just accept the default settings, leading such systems to significantly underperform relative to their potential. This problem, in turn, has impacts on cost, revenue, customer satisfaction, business reputation, and mission effectiveness. To improve the overall performance of the end-to-end systems, we propose to systematically explore (i) how to design new systems towards good performance through design space synthesis and evaluation, and (ii) how to auto-configure an existing system to obtain better performance through heuristic configuration space search. In addition, this research further studies execution traces of a system to predict runtime performance under new configurations.