Mapping IP QoS to Underlying Transmission Networks
IP is traditionally able to run over almost any network transmission layer. While this approach has a clear advantage in terms of connectivity and interoperability on the network layer, it also means that only common-denominator functionality is required from the transmission layer. Even if the underlying transmission protocols offer advanced services, such as fragmenting or reassembly of packets and QoS-handling capabilities, IP standardization has a tendency to be self-sustained and either ignores these capabilities or redesigns them within the network or transport layer.
Along these lines, IP QoS elements have been designed as standalone mechanisms. Any mapping onto lower-layer QoS elements, such as ATM traffic classes or Multi-Protocol Label Switching (MPLS) traffic engineering attributes, are either not foreseen or are in the early stages of design and implementation (often driven by interested vendors).
While mapping IP QoS elements to fixed-quality transmission services such as ATM traffic classes may seem straightforward, the dynamic mapping of IP QoS to MPLS attributes is a much more difficult task. For this mapping, path-forwarding state and traffic management or QoS state is defined for traffic streams on each router along the path. According to the general model in RFC 2475:
“...traffic aggregates of varying granularity are associated with a label switched path at the ingress point, and packets/cells within each label ...