Container-Based Virtual Labs for Scalable Cloud-Based Education

Abstract

Container-based virtual laboratories leverage lightweight operating-system virtualization technologies—chiefly Docker containers orchestrated via Kubernetes—to provide flexible, reproducible, and cost-effective practical environments for cloud-based education. This study presents the design, deployment, and evaluation of a containerized lab framework tailored to undergraduate computer science curricula, comparing traditional VM-based labs with containerized solutions (with and without autoscaling). A total of 120 students across three cohorts completed identical eight-week modules on programming and networking. Key performance indicators included container startup latency, resource utilization, task completion time, and student satisfaction. Containers booted in an average of 4.2 s (SD = 0.8), versus VM boot times exceeding 90 s.

Autoscaling maintained CPU utilization at 65% (SD = 10%), avoiding the peaks (78%, SD = 12%) seen in non-autoscaled setups. Students in the autoscaled container cohort completed assignments 20.9% faster (M = 25.4 min, SD = 4.2) than those using VMs (M = 32.1 min, SD = 6.5; t(78) = 7.45, p < 0.001) and reported higher satisfaction (M = 4.3/5). These findings demonstrate that container-based labs with dynamic scaling dramatically improve provisioning speed, resource efficiency, and learning outcomes, while reducing infrastructure overhead. The paper concludes with best-practice recommendations and discusses scope and limitations.