Transport Networks

Transport networks are different than IP and other service networks.  I think of transport in terms of a simple equation:

Transport = Transmission + Switching

Transmission

Transmission is the component of Transport systems that is associated with distance, and operates at the physical layer.

Transport networks have the following functional requirements:

1. Support for Point-to-point connections
2. Traffic Engineering and admission control for deterministic performance
3. Traffic Separation
4. OAM for proactive detection and isolation of failures
5. Protection and Service Restoration mechanisms 
6. Timing and Synchronization support

Transport networks also have a number of non-functional requirements:

1. High Availability
2. Low Cost
3. Scalability
4. Efficiency
5. Security

The per-port cost of transport equipment is significantly lower than the per-port cost of routers. Adding these transport network functions to a service switching platform like a router, adds to their cost. A network that focuses only on transport requirements, can meet the transport requirements at significantly lower cost than can routers that have been enhanced to support transport functions. Cost is reduced if the more expensive routing functions are deployed only at the ends of point-to-point transport tunnels.

A growing part of some carrier's revenue is from wholesale transport service, and almost all carriers rely on other carriers to provide them with wholesale transport services. These do-it-yourself customers want bandwidth pipes, not IP services: The services layer functions can be eliminated from the transit nodes when transporting wholesale services. Eliminating these functions reduces state which saves both capital costs and operating costs.

Transport technology is inherently multiservice. The carrier should be free to transport any services he chooses. And the unique attributes of the different services shouldn't require changes to the transport equipment. An optical transport network that transports SONET/SDH today, should be able to transport Ethernet or MPLS tomorrow.

Transport is the foundation of the network, so it must be reliable. Simplicity is key to achieving reliability. Human error accounts for a significant number of network outages. Keeping things simple reduces the chance of human error. The transport network must support simple service turn-up and maintenance procedures that are consistent, and independent of the service type, or the transport technology.

A friend at a large tier-1 carrier tells me that the majority of their operating costs are associated with the physical network. OAM that is capable of identifying and isolating faults in the physical network reduces these costs. The requirements of physical OAM are a superset of the configuration and equipment focused OAM used at the higher network layers: transport OAM must detect physical faults, and quickly trigger protection switching. Unlike the OAM that is commonly used in IP and MPLS networks, transport OAM targets fully automated fault isolation and root cause analysis. Transport OAM is part of a proactive network monitoring and automatic diagnostic function. Contrast this with reactive OAM tools, like Ping and TraceRoute. These are great tools for experts to use to manually diagnose network configuration and performance problems at the IP layer, but are slow and expensive when diagnosing problems in the transmission network.

Switching

Connections in transport networks are setup and remain in service for long periods of time. Transport networks are expected to provide a stable foundation for the network. Connections operate at much higher bandwidth and consume expensive  resources. Transport networks have a number of unique switching requirements.

1. Switching in transport network occurs at multiple levels and uses multiple technologies
2. Each technology layer, from the duct or conduit, on up to the service aggregation layers, has its own network topology.
3. Cost constraints demand that path computation factor in Traffic Engineering requirements
4. Transport costs are crtically dependent on traffic aggregation. Aggregation includes Grooming within each layer, and multiplexing between the layers
5. This gets complex fast. Software control of transport switching functions is a critiical requirement.
6. Manageability demands non-blocking switching in all layers.

The lowest swtiching layers provide capacity. Higher layers groom and aggregate at finer granularity.  The higher layers have more associated state information, and require more complex switching hardware and software. The lower switching layers include fiber and fixed DWDM. Next comes ROADM and OTN. SONET/SDH and Connection Oriented Ethernet provide the fine grained transport switching functions at the top of the transport stack.

The bottom line is that transport networks have a distinct set of requirements. The required capabilities are best served by equipment and software that is optimized for transport applications.