MATLAB Production Server Hot-Deployment Memory Limits

MATLAB Production Server (MPS) supports hot-deployment, allowing updated compiled archives to be uploaded without restarting the server instance. This mechanism enables zero-downtime updates by routing requests to the newest available version while existing workers complete their current tasks.

A potential constraint arises during the transition period when multiple versions of the same archive coexist in the worker memory. For applications with large memory footprints or extensive static data loading, this overlap may lead to unexpected resource exhaustion before the older version is fully purged from the worker pool.

• How does MPS manage the memory ceiling when transitioning between two large archives on a single worker?
• Is there a configurable threshold to trigger the immediate release of the previous archive version once the new version is active?