Skip to main content

Deployment overview

FLNet is designed as a controlled platform, which means deployment is part of the product architecture rather than an afterthought.

Core deployment idea​

The platform separates:

  • the control plane, where definitions, metadata, runs, and state transitions are managed
  • the execution plane, where containerized tools actually run

This separation is important for both security and maintainability.

Typical deployment responsibilities​

A production-oriented deployment usually needs to provide:

  • authentication and authorization
  • persistent storage for metadata and artifacts
  • a backend API as the stable control layer
  • an orchestration service for runtime execution
  • a frontend for user interaction

Depending on the environment, these components may be deployed together or split across services.

Why deployment quality matters​

From the perspective of the paper, reproducibility is not only about code. It also depends on:

  • stable execution environments
  • controlled container images
  • explicit versioning
  • persistent run state
  • observable status transitions

A weak deployment can undermine all of those properties.

Treat deployment as support for controlled scientific execution, not just app hosting.

That means you should care about:

  • environment consistency
  • traceable configuration
  • separation of privileges
  • monitoring and logs
  • safe update procedures