Allied Telesis AT-S63 - Manual (Page 338)

Chapter 17: Quality of Service
338 Section II: Advanced Operations
be added to another traffic class. A traffic class may have many flow
groups. Traffic is matched in the order of the flow groups. For example, if a
traffic class has flow groups 1, 3, 2 and 5, this is the order in which the
packets are matched.
QoS controls at the flow group level provide a QoS hierarchy. Non-default
flow group settings are always used, but if no setting is specified for a flow
group, the flow group uses the settings for the traffic class to which it
belongs. For example, you can use a traffic class to limit the bandwidth
available to web and FTP traffic combined. Within that traffic class, you
can create two different flow groups with different priorities, to give web
traffic a higher priority than FTP. Web traffic would then be given
preferential access to bandwidth, but would be limited to the bandwidth
limit of the traffic class.
Traffic Classes
Traffic classes are the central component of the QoS solution. They
provide most of the QoS controls that allow a QoS solution to be deployed.
A traffic class can be assigned to only one policy. Traffic classes consist of
a set of QoS parameters and a group of QoS flow groups. Traffic can be
prioritized, marked (IP TOS or DSCP field set), and bandwidth limited.
Traffic is matched in the order of traffic class. For example, if a policy has
traffic classes 1, 3, 2 and 5, this is the order in which the packets are
matched.
Policies
QoS policies consist of a collection of user defined traffic classes. A policy
can be assigned to more than one port, but a port may only have one
policy.
Note that the switch can only perform error checking of parameters and
parameter values for the policy and its traffic classes and flow groups
when the policy is set on a port.
QoS controls are applied to ingress traffic on ports. Therefore, to control a
particular type of traffic, an appropriate QoS policy must be attached to
each port that type of traffic ingresses.
Although a policy can be applied to an egress port, the classifiers and the
QoS controls are actually applied by the switch on the ingress ports of the
traffic. This means the parameters used to classify the traffic and the
actions specified by the policy are checked and applied on the ingress
traffic of every port, before the traffic reaches an egress queue. As a
consequence, a policy is never applied to the whole aggregated traffic of a
designated egress port, but rather to the individual ingress flows destined
to the port.
The effects of this behavior become evident when using the maximum
bandwidth feature of QoS. Here is an example. Suppose you have a policy
that assigns 5 Mbps of maximum bandwidth to an egress port. Now
assume there are 10 ports on the switch where ingress traffic matches the
criteria specified in the classifier assigned to the policy of the egress port.