Abstract

The process of building and deploying a commercial software product generates a diverse set of data artifacts. Examining these artifacts provides insight into a team’s development process and sheds light on customers’ experiences using production versions in the field. However, before we can leverage these diverse data sources to ask questions, solve problems, and improve these products, we have to dig into and understand the data we are collecting so that we can extract useful information. This workshop will focus on two case studies involving data collected during the development of, and operation of, a commercial statistical software product. During the discussion of each case study, we will focus on highlighting the problem to be solved, and we will then discuss the technologies that generate and process the data we are using to try to solve it. Our development team faces challenges that are similar to the range of challenges that other product development teams working in commercial software development face, and the approaches we take to explore our data are not unique. Any number of data acquisition and analysis tools, along with a suitable programming language, could be used to tackle these tasks, though we choose to use our own product. The first case study involves automation of access and parsing of machine-generated text log data from Jenkins-based build system projects for pre-production software to create structured tables for analysis and visualization of performance, build failure and test details. The second case study focuses on how our crash reporting system has transformed the way we collect and analyze information on software crashes from testing teams as well as external and internal users.

About the Speakers

Shannon Conners has managed production of new software releases within the research and development group of the JMP division of SAS since 2012. Before joining JMP in 2006 as a Genomics Product Manager, she used SAS and JMP to analyze and visualize microarray data from microbes from extreme environments during her PhD studies and postdoctoral work in bioinformatics at North Carolina State University.

Joseph Morgan has worked with a wide cross-section of JMP scripting users to provide guidance on developing efficient, robust, production-ready applications over the past 10 years. Before joining JMP, he was a full-time faculty member in the School of Computer Science at DePaul University, where he taught software engineering and data analysis. He is published in many journals and conference proceedings; his research interests span combinatorial design methods to empirical software engineering.

[Back]