26-27
Catalyst 3750 Metro Switch Software Configuration Guide
78-15870-01
Chapter 26 Configuring QoS
Understanding Hierarchical QoS
Hierarchical Queues
The switch uses a hierarchical queueing model for traffic sent from an ES port. Each packet is assigned
a queue based on its physical interface, VLAN, or class:
• At the class level, a packet is queued to one of four queues according to its CoS, DSCP, IP
precedence, or MPLS EXP classification. Packets can be classified by any combination of these
values, but if a packet matches more than one, the classification occurs in the order listed. The last
queue in each set of four queues is reserved as the default queue. Packets that are not classified into
one of the other three queues are assigned to the default queue. You can configure traffic in the
default queue with congestion-avoidance features and scheduling congestion-management features
as described in the “Queueing and Scheduling of Hierarchical Queues” section on page 26-26.
• At the VLAN level, the switch supports 2045 VLAN classes divided between the two ES ports. One
queue is reserved as the default queue. Packets that are not classified into one of the other VLAN
queues are assigned to the default queue. You can configure traffic in the default queue with
congestion-avoidance features and scheduling congestion-management features as described in the
“Queueing and Scheduling of Hierarchical Queues” section on page 26-26.
• At the physical level, the switch reserves one queue per port.
The switch creates the default queue and uses it to send all traffic when an egress service policy is not
attached to an ES port. User traffic on a physical port without an attached service policy bypasses the
QoS classification and is queued to the default queue. The minimum and maximum bandwidth for the
default queue is the same as the port bandwidth.
Under congested conditions, the switch discards packets for all classes configured for the same sending
queue with equal probability. To achieve the full queueing capacity, there must be an equal division of
traffic among the classes for each sending queue.
Congestion-Management and Congestion-Avoidance Features
You use congestion-management features to control congestion and to control the order in which
outbound packets are sent from an ES port based on the priorities assigned to those packets. You manage
congestion by creating queues, by assigning packets based on the packet classification, and by
scheduling the packets to be sent from the queue.
During periods with light traffic (when no congestion exists), the switch sends packets as soon as they
arrive. During periods of congestion at the outbound port, packets arrive faster than the port can send
them. If you use congestion-management features, the switch queues accumulating packets at a port until
it is free to send them. They are then scheduled for transmission according to their assigned priorities
and the queueing mechanism for the port.
You can configure either tail drop or WRED. You cannot configure both tail drop and WRED in the same
class policy, but they can be used in two different class policies in the same policy map.
You can configure CBWFQ as a queue scheduling management feature, LLQ as a scheduling
congestion-management feature, and traffic shaping to decrease the burstiness of traffic.
Tail Drop
With tail drop, packets are queued for the class until the maximum threshold is exceeded, and then all
the packets destined for the class queue are dropped. You enable tail drop at the class level by using the
queue-limit policy-map class configuration command. For configuration information, see the
“Configuring an Egress Hierarchical QoS Policy” section on page 26-77.
To Next Page
Loading...
Need help?
General
