Coding in the Cloud - Capturing Programming Behaviors at Scale

Access to computing resources, specifically highperformance parallel systems, is still a fundamental barrier to providing the necessary learning environment for many students in the Parallel and Distributed Computing (PDC) courses. Many top-ranked universities may have ample computing facilities for both research and education, however, other programs are not able to provide competitive infrastructure and therefore must develop different strategies. While virtual machine environments on student-owned personal computers are effective for providing a consistent low-cost and uniform platform to students, there are deficiencies in their use for specific areas of the PDC curriculum, such as performance analysis, tuning, and optimizations techniques. Moreover, the virtualized bring-your-own-system model eliminates a core and critical opportunity emerging in cloud environments, that development in the cloud can be centrally analyzed by teachers to assess learning objectives, evaluate course outcomes, and provide dynamic feedback to instructors. Forms of cloud coding in which programmers develop and run on remote servers using web-based interfaces have significant potential in broadening teaching efforts in PDC. The talk will present the experience with providing a large class of upper-level undergraduate and graduate students with access to a high-performance cloud coding server that supports IPython Notebooks. In addition to the principles of accessing the system from any web browser, the developed framework allows instructors to observe, replay, and catalog the student programming process. The system has been used to help integrate multi-threading and GPU programming.