Component redundancy is critical to reduce MTTR for the system and primarily consists of the following router features:
Dual route processor modules — For a highly available architecture, redundant Control Processor Modules (CPM) are essential. The route processing functions of the CPM calculate the most efficient route to an Internet destination and communicate the best path information to peer routers. Rapid information synchronization between the primary and secondary CPMs is crucial to minimize recovery time.
Switch fabric (SFM) redundancy — Failure of a single switch fabric card with little to no loss of traffic.
Redundant line cards — LAG, ECMP and other techniques to spread traffic over multiple line cards so that a failure of one line card does not impact the services being delivered.
Redundant power supply — A power module can be removed without impact on traffic.
Redundant fan — Failure of a fan module without impacting traffic.
Hot swap — Components in a live system can be replaced or become active without taking the system down or affecting traffic flow to/from other modules.
Router hardware architecture plays a key role in the availability of the system. The principle router architecture styles are centralized and distributed. In these architectures, both active and standby route processors, I/O modules (IOMs) (also called line cards), fans, and power supplies maintain a low MTTR for the routing system.
However, in a centralized architecture, packet processing and forwarding is performed in a central shared route processor and the individual line cards are relatively simple. The cards rely solely on the route processor for routing and forwarding intelligence and, should the centralized route processor fail, there is greater impact to the system overall, as all routing and packet forwarding stops.
In a distributed system, the packet forwarding functionality is situated on each line card. Distributing the forwarding engines off the central route processor and positioning one on each line card lowers the impact of route processor failure as the line cards can continue to forward traffic during an outage.
The distributed system is better suited to enable the convergence of business critical services such as real-time voice (VoIP), Video, and VPN applications over IP networks with superior performance and scalability. The centralized architecture can be prone to performance bottleneck issues and limits service offerings through poor scalability which may lead to customer and service SLA violations.