Carrier Core Router Product

Overview

A carrier-grade core router is the backbone switching element in a service provider network, handling the highest traffic loads and linking together the entire internet. These devices are built for extreme availability: they forward terabits of traffic while remaining up during component failures, maintenance, and configuration changes. The architecture is modular—line cards slide into slots on a central chassis, all connected via a high-capacity switch fabric. The Chassis and Midplane is the skeleton, the Switch Fabric is the traffic highway, [[core-router-route-processor|route control cards]] run the routing protocols, and [[core-router-line-cards|line cards]] are the ingress/egress endpoints that connect the network segments.

How it works

The Route Control Card cards run BGP (Border Gateway Protocol) and other routing protocols, learning paths through the internet and building a forwarding table (FIB). Each [[core-router-line-cards|line card]] has a local copy of this table and an on-board [[core-router-lc-forwarding-chip|forwarding ASIC]]. When a packet arrives on a port (via a 10/40G PHY), the ASIC performs a table lookup and decides which outbound port to forward to—potentially on another line card. The packet travels across the Switch Fabric via high-speed SerDes links, then exits via another line card's port.

The routing fabric's bandwidth is the router's limiting throughput. With a 100 Tbps fabric, the router can sustain wire-rate forwarding on all ports in aggregate; a 10G port saturating ingress and egress simultaneously consumes negligible fabric bandwidth relative to the whole. Packets arriving faster than they can be forwarded are buffered in the [[core-router-lc-buffer|packet memory]]; if buffers overflow, packets are dropped (tail-drop), a signal to TCP that the network is congested.

Redundancy is built into every subsystem. The [[core-router-route-processor|control cards]] are N+1, so if one fails, the standby takes over without a routing protocol restart. [[core-router-power-system|Power supplies]] are redundant, and the [[core-router-cooling|fan subsystem]] can lose one tray without thermal shutdown. Line cards can be removed and replaced without bringing down the whole router; traffic on that line card is dropped, but all other forwarding continues. This modularity also means a carrier can start with a small configuration (e.g., 6 line cards) and add more over years without replacing the entire box.

Physical organization

The [[core-router-chassis|chassis frame]] provides the structure. The [[core-router-midplane|midplane]] is the copper or aluminum backplane PCB that carries data and control buses to every slot. The [[core-router-slot-cage|slot cage]] guides hot-swappable cards in and out with precision. A [[core-router-thermal-wall|thermal baffle]] inside separates hot exhaust air on one side from cool intake on the other, ensuring efficient convective cooling. The [[core-router-cooling|fan system]] pulls cool air through the intake side, absorbs heat from the components, and pushes hot air out the rear.

Redundancy and resilience

Modern core routers are designed for five-nines availability (99.999% uptime), achievable only through N+1 or N+2 redundancy and careful engineering. The [[core-router-route-processor|route processor cards]] run a synchronization protocol (often called "in-service software upgrade" or ISSU) that keeps the standby card's routing table and BGP session state in sync with the active card. If the active card fails or is removed for maintenance, the standby seamlessly takes over—possibly with a few seconds of microloop or packet loss, but without a full routing protocol restart that would flap upstream links.

The [[core-router-fabric-chips|fabric itself]] is often redundant too, with N+1 or even full mesh interconnect, so a single chip failure does not partition the forwarding plane. Power and cooling redundancy mean no single PSU or fan failure stops the router. Line card failures are localized; only traffic on that card is affected, and downstream neighbors notice loss of signal on those circuits.

Scaling and evolution

Core routers are engineered to grow. Newer line cards with faster optics (e.g., 400G) can be inserted into older chassis, as long as the fabric bandwidth supports them. Control plane capacity—how many routes the [[core-router-route-processor|route processor]] can hold—is limited by memory, but gigabyte-scale DRAM is now standard. A router holding 2 million BGP routes and 10 million FIB entries is common in carrier backbones.

The [[core-router-raid-controller|RAID controller]] (if present) manages any local disk storage for logging or software images, often using RAID-5 or RAID-6 for fault tolerance.