Unlocking Optimal ORM Database Designs: Accelerated Tradeoff Analysis with Transformers

Optimizing object-relational database mapping (ORM) design is essential for software system performance. However, current ORM design tools widely used in practice offer limited support for evaluating performance implications. State-of-the-art relying on systematic tradeoff analysis provides greater assistance but becomes impractical for large-scale, real-world systems. In this paper, we introduce an innovative solution grounded in machine learning to efficiently and scalably identify optimal ORM database tradeoffs.Our approach involves training a transformer model using a dataset of formally analyzed object-relational database designs to identify Pareto-optimal design tradeoffs, significantly reducing the number of design candidates. The extensive experiments across various software database systems demonstrate the high effectiveness of the approach in identifying optimal design alternatives overlooked by leading ORM tools. Additionally, our results show a remarkable 98.21 % improvement in analysis efficiency compared to the current state-of-the-art, reducing tradeoff analysis time from over 15 days to around 18 minutes on average.