The Differentiated Services Model 688

The Differentiated Services Model
The QoS implementation in Catalyst switches is based on the Differentiated Services
(DiffServe) architecture. This reference model states that packets are marked (classified) at
the entry point into the network, and that every subsequent router or switch, implementing
hop-by-hop forwarding, uses the classification to try to match the forwarding process to the
classification. This is achieved by each DiffServe router in the path having a locally configured
queuing priority for forwarding marked packets. Non-DiffServe-enabled routers will
simply forward packets based upon default queues. Figure 20.6 shows the DiffServe architecture,
with routers in the end-to-end path either being in the domain or without. The entry
point to the domain is called the ingress, and the exit point is called the egress.
At layer 3, this classification and marking is established by setting bits in the IP Type of Service
(TOS) field to differing values. At layer 2, however, this is a little more difficult, because
there are no fields inside legacy Ethernet available for this purpose. Even so, there are some
clever mechanisms that allow us to map layer 2 priorities to layer 3.
The basic QoS model underlying all efforts is closely related to the DiffServe architecture.
Shown in Figure 20.7, it consists of a series of discrete stages. First, the packets are classified and
tested to see if they conform to the configured classification. This stage is called policing.