Abstract
Traditional undergraduate software engineering (SE) education often emphasizes theoretical knowledge at the expense of hands-on technical proficiency, leading to a disconnect between academic training and industry requirements. This paper presents a redesigned SE course at the University of Jinan that utilizes virtualization technology to implement a project-driven, "learning-by-doing" pedagogy. By integrating a "pay-as-you-go" economic model and scaffolded project complexity, the course aims to develop both technical skills and resource management awareness. Our findings indicate significant improvements in student skill acquisition, resource efficiency, and engagement compared to traditional teaching methods.
The "Ivory Tower" Problem in SE Education
A persistent challenge in software engineering education is the "isolation of practice." Students often master theoretical concepts—referred to as "knowledge points"—but struggle to apply them in collaborative, iterative, and resource-constrained environments. Furthermore, typical lab settings provide "free" computational resources, which fails to prepare students for the economic realities of modern cloud deployment and the importance of resource optimization.
Methodology: Virtualization and Economic Constraints
To address these gaps, the course was redesigned around a custom Virtualization Technology (VT) platform. This platform served two primary purposes: providing a consistent development environment and simulating industry cost constraints.
The Economic Model
To combat resource blindness, the researchers implemented a "pay-as-you-go" billing system. Students were allocated a virtual budget of CNY 100 and charged for various actions on the platform, including VM creation, uptime, and high CPU utilization. This model forced students to consider "green computing" and efficiency, discouraging them from leaving idle instances running.
Scaffolded Project Complexity
The curriculum transitioned from a single large project to a three-stage "step-by-step" approach to manage cognitive load and collaborative friction:
- Single Project: Focused on individual technical skills such as unit testing and performance scripting.
- Pair Project: Introduced collaborative elements like role assignment and peer code reviews.
- Agile Team Project: A 3-5 person group effort utilizing rapid iterative processes (requirement analysis, design, development, and testing).
Key Results and Outcomes
The shift to a project-driven model resulted in a dramatic increase in student technical output and self-reported skill growth. Quantitative data from the 2012–2014 cohorts demonstrates the impact of the "learning-by-doing" approach on student productivity.
Skill Acquisition and Efficiency
In the 2014 cohort, student self-evaluations showed marked improvement across all domains. For instance, proficiency in "Testing" rose from a pre-course score of 2.4 to 4.8, while "Requirement Analysis" rose from 2.7 to 4.9. Furthermore, the refined 2014 curriculum proved more resource-efficient; the average cost per student dropped from CNY 98 in 2013 to CNY 83 in 2014, despite increased assignment complexity.
Student Sentiment
Engagement levels were high, with 70% of students identifying the "Agile Team Project" as the most important component of the course. Additionally, 65% of students recommended keeping the current lab and project structure unchanged, validating the pedagogical shift.
Discussion and Lessons Learned
Several key insights emerged from the implementation:
- Cost as Pedagogy: Making computing costs visible significantly altered student behavior, encouraging optimized code to avoid high-CPU billing tiers.
- The "Checking Point" Effect: Students tended to delay progress until deadlines. The researchers recommend bi-weekly checkpoints to ensure consistent effort throughout the semester.
- The Implementation Gap: A 2013 finding showed that while 93.6% of students passed theoretical design reports, only 42.1% passed functional test cases. This underscored the necessity of the hands-on model to move beyond mechanical memorization.
Conclusion and Future Work
The project-driven, learning-by-doing method effectively bridges the gap between SE theory and practice. While challenges remain—such as network stability for remote access and a tendency for some students to "reset" rather than troubleshoot—the model provides a robust framework for preparing graduates for the software industry. Future iterations will focus on increasing checkpoint frequency and adding discussion-based hours to address specific technical hurdles.
Summary updated