Parallel model validation with epsilon

Traditional model management programs, such as transformations, often perform poorly when dealing with very large models. Although many such programs are inherently parallelisable, the execution engines of popular model management languages were not designed for concurrency. We propose a scalable data and rule-parallel solution for an established and feature-rich model validation language (EVL). We highlight the challenges encountered with retro-fitting concurrency support and our solutions to these challenges. We evaluate the correctness of our implementation through rigorous automated tests. Our results show up to linear performance improvements with more threads and larger models, with significantly faster execution compared to interpreted OCL.