This practical experience report is based on systems courses at Columbia University. The authors explain that dedicated computing resources, such as course server farms and time-sharing systems, are no longer necessary for many practical computer science courses. As they note, most students have very capable laptops that run virtualization services, such as VMware, reasonably well. Courses at Columbia provide basic minimal images that students can download and install on their own machines. The students can then modify the images.

While virtual images allow each student to work at home, there remains the challenge of submitting these changes to the course administrators. A Git distributed version control system is used to collect the changes. Each student creates a branch, and all of the updates are permanently logged. The administrators can view submission logs to determine when each update was submitted, which allows them to ignore any changes made after the submission deadline.

Interestingly, even with all of these tools, the courses found it necessary to use live demonstrations as the final proof that students had actually performed the exercises and understood the material. In addition, the use of the students’ laptops for the demonstrations reduced the load on the graders, because each project did not need to be recompiled and installed on the graders’ systems.

Read this paper if you run a systems course, programming or otherwise. The use of virtualized images is a major improvement over dedicated managed hardware and software systems.