Machine learning
Optimizing Performance Testing at TEC University with Novacomp’s AI-Powered Solution
Solution: AI/ML
The Challenge:
TEC University, a prominent public educational institution, faced a critical challenge in preparing for the launch of their new tuition system. The institution needed to ensure the system’s resilience and performance under peak loads during the enrollment period. They sought to accurately model and simulate user concurrency patterns to conduct automated performance tests. This was crucial to guarantee a seamless experience for students and administrators, even during periods of high demand.
The solution:
Novacomp, in collaboration with our Agile Team Solution and IT Solutions Architect, embarked on a mission to address TEC University’s challenge. We assembled a team of Performance Testing and AI/ML experts to design and implement a cutting-edge Data Analytics and AI/ML solution aimed at creating and maintaining an accurate user concurrency/load model.
How we did it:
- Data Collection and Preparation: We began by selecting relevant data from web analytics tools, including Google Analytics. This data was then meticulously cleaned, constructed, integrated, formatted, standardized, and normalized.
- Model Development: Our team developed custom and specific linear models tailored to predict various performance targets. These models covered metrics such as the number of sessions over time and page load times.
- Validation and Adjustment: We rigorously validated these models, ensuring that the simulated concurrent access behavior closely matched historical data. The models were fine-tuned to produce realistic predictions of future behavior.
- Model Application on Performance Simulation Tests: Once validated, we configured performance simulations based on the model properties. Multiple rounds of performance testing were conducted to assess the system’s performance against acceptance criteria.
Results:
- Expected Concurrency Load Model: TEC University now has accurate models for system behavior during peak times.
- More Accurate and Flexible Representation: Dynamic models adapt to new data, providing a precise representation of system behavior.
- Projection and Predictive Ability: Models can simulate various scenarios by adjusting parameters.
- Formal, Simple, Versatile, and Reusable: A straightforward linear model informs automated performance tests.
