TAU Commander: An Intuitive Interface for the TAU Performance Analysis System

The Department of Energy and other federal agencies have made significant investments in software performance engineering tools, which can be complex and difficult to use. Novice users first encountering a tools complexity and vast array of features are intimidated and easily frustrated. There is also a lack of advanced problem identification capabilities so users rely heavily on their own experience and intuition to identify the root cause of a software performance problem. These problems limit the tools adoption by small companies and independent software vendors, particularly in the manufacturing sector. This project is developing a complete production grade software performance engineering product that lowers the barriers to entry for novice users and enhances their ability to mine actionable information from software performance data. The new product presents a simple, intuitive, and systemized user interface that guides users through performance engineering workflows and will use advanced cloud-hosted performance analysis services to offer unprecedented data analysis and problem detection capabilities. Phase I demonstrated the feasibility of the approach by developing a prototype performance engineering product that successfully resolved pain points in every stage of the performance engineering workflow and measurably simplified tool usage. In a study of 124 unique workflows, the prototype shortened workflows by approximately 50% and reduced the number of commands a user must master from eight to exactly one. Phase II will further develop the new product, particularly for applications that use co-processing accelerators, and demonstrate the feasibility of using cloud-hosted performance analysis services to improve productivity. Cloud-hosted performance data storage, analysis, reporting, and visualization services will be prototyped using the new product as client-side software. Commercial Applications and Other Benefits: Performance engineering is critical to delivering the full potential of high-end parallel computing, a fundamental technology for science and engineering discovery. We anticipate that products developed in part by this project will improve engineering capabilities for approximately half of the roughly 280,000 manufacturers in the United States who missed deadlines or restricted product development in 2012 due to performance limitations of on-site computing resources.