Enterprise Deployment Guide

Tune Rescile's stabilization loop, memory usage, and generator caching for large enterprise graph estates.

Edit this page on GitHub

Performance & Scaling

Rescile’s graph engine is designed to be highly performant, but extremely large enterprise estates require tuning.

The Stabilization Loop

The model engine runs in a stabilization loop, processing models until no new resources or properties are created. By default, this is capped at 8 iterations.

If your graph has deeply nested derivations (e.g., A derives B derives C derives D…), you may hit this limit prematurely. You can increase it via the CLI:

rescile-importer --max-loop 12

Large Graph Considerations

  • Memory: The graph is built entirely in memory. Ensure your builder node has sufficient RAM (typically 2-4GB is plenty for estates up to 100,000 nodes). The stateless Controller fleet is highly memory efficient.
  • Targeted Outputs: If output generation is slow, ensure your match_on filters are specific, so you aren’t needlessly generating artifacts for every single node.
  • Generators: Rely heavily on ttl caching and condition = "on_missing" for heavy API-polling generators.