Storage researchers have always been interested in understanding the complex behavior of storage systems with the help of statistics, machine learning, and simple visualization techniques. However, when a system's behavior is affected by hundreds or even thousands of factors, existing approaches break down. Results are often difficult to interpret, and it can be challenging for humans to apply domain knowledge to a complex system. We propose to enhance storage system analysis by applying "interactive visual analytics," which can address the aforementioned limitations. We have devised a suitable Interactive Configuration Explorer (ICE), and conducted several case studies on a typical storage system, to demonstrate its benefits for storage system researchers and designers. We found that ICE makes it easy to explore a large parameter space, identify critical parameters, and quickly zero in on optimal parameter settings.
