// HOW IT WORKS

A disaggregated edge, plane by plane.

Elastic BNG and eSwitch share one architecture: three independent planes that scale, upgrade, and fail on their own terms. Here's how a subscriber packet moves through it.

PACKET WALK · LIVE
3 PLANES · 1 FABRIC
PLANE 01
Management

Declarative intent, GitOps pipelines, streaming telemetry, and lifecycle. This is where operators express what the network should do — and where every change is versioned, canaried, and reversible.

PLANE 02
Control

Session state, AAA, routing, and policy computation run as a highly-available cluster. It programs the user plane over a standard interface (PFCP-style) and holds the authoritative view of every subscriber.

PLANE 03
Data

Line-rate forwarding on merchant silicon or DPDK. Stateless instances you add and remove freely — QoS, CGNAT, and encapsulation all happen here, at sub-microsecond latency.

// DESIGN PRINCIPLES

Four rules the whole platform obeys.

Scale out, never up

Capacity is a number of instances, not a hardware SKU. Every plane scales horizontally on commodity servers and whitebox switches.

State is replicated, not pinned

Subscriber state lives in a distributed store. Any data-plane instance can take over a session, so a failure drains instead of drops.

Standards at every seam

PFCP, RADIUS, BGP, gNMI, OpenConfig. Interfaces are open so the platform drops into networks you already operate.

Observable by default

Every plane streams telemetry continuously. You debug the live network with the same data the control loop uses.

Want the reference design?

Our architects will walk your team through a deployment tailored to your topology.

Talk to an architect Deployment models →
© 2026 Elastic BNG, Inc. — ebng.net
Privacy Terms Security Status