Skip to content

CONCEPT Cited by 1 source

Workload identity

A workload identity is a stable, fine-grained identifier naming the logical unit of software running on a host — typically more specific than "this instance" (AMI/VM) and as specific as "this container" or "this service". Workload identities are the canonical resolution target for concepts/ip-attribution: any given packet or flow should be attributable to the workload identity that owns it.

Provisioning

Two canonical shapes appear in Netflix's architecture:

  • EC2-instance path. A boot-time provisioning service (Metatron at Netflix) issues an identity certificate per instance; the identity is readable from local disk by any process running on the instance.
  • Container path. A per-host daemon (IPManAgent) writes IP → workload-ID (and (IP, port) → workload-ID for shared-IP scenarios) into an eBPF map when each container starts; kernel-resident observers can resolve the workload owning a socket without a userspace round-trip (patterns/ebpf-map-for-local-attribution).

Why this matters

Workload identities turn "who is at this IP?" from a topology question into a queryable fact at capture time. Local workload identity resolution is the load-bearing move that makes accurate heartbeat-based ownership possible — every flow emerges pre-tagged with the local identity, so downstream attribution doesn't need to reconstruct it.

Adjacent but distinct

Workload identity ≠ concepts/workload-identity in the IAM / SPIFFE sense (which concerns authenticating a workload to other services). The Netflix post uses the term for the attribution side: who owns the packet, not who is allowed to send it. Both problems benefit from the same substrate (the Metatron cert), but the consumer surfaces are different.

Seen in

Last updated · 542 distilled / 1,571 read