Cisco Catalyst 3750 Software Configuration Manual page 508

Understanding Hierarchical QoS
WRED
Cisco Systems implements a version of Random Early Detection (RED), called WRED, differently from
other congestion-avoidance techniques. WRED attempts to anticipate and avoid congestion, rather than
controlling congestion when it occurs. WRED takes advantage of the TCP congestion control to try to
control the average queue size by signaling end hosts when they should temporarily stop sending
packets. By randomly dropping packets before periods of high congestion, it tells the packet source to
decrease its sending rate. Assuming the packet source is using TCP, WRED tells it to decrease its sending
rate until all the packets reach their destination, meaning that the congestion is cleared. By dropping
some packets early rather than waiting until the queue is full, WRED avoids dropping large numbers of
packets at once.
When a packet arrives and WRED is enabled, these events occur:
•
•
•
•
You enable WRED by using the random-detect policy-map class configuration command at the class
level. This command allows for preferential drop treatment among packets with different IP precedence
or DSCP values. The WRED algorithm discards or marks packets destined for a queue when that queue
is congested. It discards packets fairly and before the queue is full. Packets with high IP-precedence
values are preferred over packets with low IP-precedence values. For configuration information, see the
"Configuring an Egress Hierarchical QoS Policy" section on page
CBWFQ
CBWFQ provides guaranteed bandwidth to particular traffic classes, such as voice, that are delay
sensitive, while still fairly serving all other traffic in the network. You define traffic classes based on
match criteria. Packets satisfying the match criteria for a class constitute the traffic for that class. A
queue is reserved for each class, and traffic belonging to a class is directed to the queue for that class.
The bandwidth assigned to a class is the minimum bandwidth that is delivered to the class during
congestion. CBWFQ uses the bandwidth weight to ensure that the queue for the class is serviced fairly.
You enable CBWFQ and specify the minimum bandwidth as a rate in kbps or as a percentage of the
available bandwidth by using the bandwidth policy-map class configuration command at the class level
or at the VLAN level. During periods of congestion, the classes are serviced in proportion to their
configured bandwidth. The amount of bandwidth available to a class is dependent on the amount of
bandwidth reserved by the parent class. For configuration information, see the
Hierarchical QoS Policy" section on page
Catalyst 3750 Metro Switch Software Configuration Guide
26-28
The average queue size is calculated based on the previous average and the current size of the queue.
The average queue-size calculation is affected by the exponential weight constant setting in the
random-detect exponential-weight-constant policy-map class configuration command.
If the average queue size is less than the minimum queue threshold, the arriving packet is queued.
The minimum queue threshold is configured through the min-threshold option in the random-detect
{dscp | precedence} policy-map class configuration command.
If the average queue size is between the minimum queue threshold and the maximum queue
threshold, the packet is either dropped or queued, depending on the packet-drop probability. The
packet-drop probability is based on the minimum threshold, the maximum threshold, and the
mark-probability denominator. The maximum queue threshold is configured through the
max-threshold option, and the mark-probability denominator is configured through the
mark-prob-denominator option in the random-detect dscp {dscp | precedence} policy-map class
configuration command.
If the average queue size is greater than the maximum queue threshold, the packet is automatically
dropped.
26-77.
26-77.
Chapter 26 Configuring QoS
"Configuring an Egress
78-15870-01