\begin{abstract}
As the size of software and system models grows, scalability issues in the current generation of model management languages (e.g. transformation, validation) and their supporting tooling become more prominent. To address this challenge, execution engines of model management programs need to become more efficient in their use of system resources. This paper presents an approach for partial loading of large models that reside in graph-database-backed model repositories. This approach leverages sophisticated static analysis of model management programs and auto-generation of graph (Cypher) queries to load only relevant model elements instead of naively loading the entire models into memory. Our experimental evaluation shows that our approach enables model management programs to process larger models, faster, and with a reduced memory footprint compared to the state of the art.
\end{abstract}