Skip to main content

Bring Your Own Container (BYOC)

info

Datalayer will publish AWS Marketplace offerings for BYOC-enabled deployment patterns soon.

Bring Your Own Container (BYOC) allows Datalayer Runtimes to enroll remote containers as execution targets for workloads.

These enrolled containers behave like virtual pod capacity from the user point of view: workloads are scheduled to them through the runtime control plane, while the container itself can live outside the primary Kubernetes node pool.

What BYOC Enables​

  1. Reuse pre-existing containers managed by another platform.
  2. Attach specialized runtime images without rebuilding the cluster base image.
  3. Expand workload capacity with remote container pools.
  4. Keep a consistent Datalayer runtime API for users.

Enrollment Model​

A typical BYOC enrollment flow is:

  1. Register a remote container endpoint with Datalayer Runtimes.
  2. Publish capability metadata (CPU/GPU/memory/runtime type/labels).
  3. Establish trust and access policy (identity, tokens, network allow rules).
  4. Expose health and heartbeat so the scheduler can place workloads safely.
  5. Mark the container as available for specific workload classes.

Architecture Behavior​

After enrollment, Datalayer can treat remote containers as runtime slots similar to virtual pods:

  1. Workloads are assigned through runtime scheduling policy.
  2. Isolation and limits are enforced per container target.
  3. Runtime status is surfaced through the same operational interfaces.

Operational Notes​

  1. Use BYOC for controlled remote capacity, not as an unmanaged bypass.
  2. Restrict enrollment to trusted container endpoints.
  3. Define clear lifecycle ownership for image updates and container rotation.
  4. Monitor heartbeat and readiness to avoid stale capacity.