In Apstra 5.1, the worker node's primary purpose is to add scalable runtime capacity to an Apstra cluster by hosting off-box services that would otherwise consume resources on the controller. Specifically, worker nodes run containerized services such as off-box device agents (used to communicate with and manage devices) and Intent-Based Analytics (IBA) components (such as probes and analytics-related services). This design keeps the controller node focused on cluster management and control-plane functions (API handling, cluster-wide state, blueprint control workflows), while shifting resource-intensive operational services to worker nodes.
As your fabric grows-more switches, more telemetry, more devices requiring agent connectivity-CPU and memory demand increases notably, especially when IBA is enabled. Adding worker nodes allows you to scale those container workloads horizontally without redesigning the fabric or reducing analytics coverage. In a Juniper data center built on EVPN-VXLAN with Junos v24.4 leaf-spine roles, this separation helps ensure that Apstra can continuously validate intent, process streaming telemetry, and maintain device communications reliably at scale. Worker nodes therefore exist primarily to offload and scale operational agents and IBA services, improving performance and resilience for larger deployments.