liveBook · Manning

Control plane

As explained in Designing AI Systems

A control plane is the part of a distributed system responsible for deployment, configuration, operations, and management rather than handling application requests.

A control plane manages a distributed application's deployment and operation; a separate data plane handles application requests. In a workflow platform, the control plane packages code, creates container replicas, registers routes, monitors health, and scales workloads, while SDK runtime servers execute inside those containers. section 9.0 section 9.8

The separation assigns responsibilities according to when and where they must run:

  • Deployment-time concerns include packaging workflow code, creating Kubernetes deployments, registering API Gateway routes, and configuring autoscaling. section 10.12
  • Request-time concerns remain in the data plane: containers receive HTTP requests, invoke workflow functions, and coordinate platform-service calls. section 9.0
  • Synchronous and streaming requests bypass the control plane, while asynchronous workflows call it to create and manage job state. section 9.0
  • Scaling and resource settings flow through deployment; timeouts and retry behavior remain enforced inside the container runtime. section 9.8

Scaling an asynchronous research workflow

A research workflow can return a job ID immediately, fan out searches across several child workflows, and later synthesize their results. The control plane independently scales those stateless containers, while sessions, memory, credentials, and other state remain in platform services. section 10.12

Figure 9.1 Control plane vs. data plane. On the left, the Workflow Service (control plane) connects to Kubernetes to manage container Deployments and to the API Gateway to register routes. On the right, the data plane: multiple workflow containers, each running the SDK's runtime server, receiving HTTP requests from the API Gateway and making gRPC calls to platform services through the gateway. For synchronous and streaming workflows, the control plane and data plane do not communicate during request handling. For asynchronous workflows, the data plane calls the control plane to manage job state.

Trace deployment and request traffic across the control plane, data-plane containers, Kubernetes, API Gateway, and platform services. section 9.0

How Designing AI Systems covers it

How deep it goes
Explains responsibilities, request-path boundaries, configuration flow, and asynchronous exceptions without detailing control-plane APIs.
Examples
Uses workflow containers, route registration, replica bounds, asynchronous job IDs, and parallel research sub-workflows.
What it assumes
Assumes familiarity with distributed services, containers, HTTP requests, and platform-managed state.
What you can do after
Operate/configure workflow deployments by separating platform management, request execution, scaling, resources, and runtime reliability.
How it connects
Connects deployment pipelines, data planes, API gateways, Kubernetes, autoscaling, stateless containers, and asynchronous workflows.

Also explained in

Compare how 10 books explain control plane →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Designing AI Systems. The section links point to the parts of the book it draws on.