Skip to content

CONCEPT Cited by 1 source

Workload federation

Workload federation is the pattern of presenting a single logical endpoint to workload submitters while routing work across multiple underlying clusters (cells) transparently.

Definition

In a federated compute system, callers submit workloads and manage capacity through one API endpoint. The federation layer handles routing, replication, and topology awareness across the underlying cluster fleet. This decouples users from cluster topology changes — cells can be added, removed, or rebalanced without client-side changes.

At Netflix (Titus)

Titus provides workload federation across multiple Kubernetes cells and federated capacity reservations. Compute Managed Batch (and its successor Netflix Batch) talks to a single Titus endpoint for job submission and capacity reservation without awareness of the underlying cell topology. Netflix Batch adds a custom Kueue router in the Titus federation layer that routes jobs to Kueue-enabled cells.

(Source: sources/2026-06-22-netflix-how-netflix-simplified-batch-compute-with-kueue)

Seen in

Last updated · 559 distilled / 1,651 read