Chapter 1 Configuring Class-Based Queuing
As an extension of WFQ, class based queuing (CBQ) provides users with class
definition support. CBQ assigns individual FIFO reservation queues to the classes
defined by each user to buffer data of the same class. When there is network
congestion, CBQ matches outbound packets according to the classification rule
defined by users to make them enter relevant queues. Before queue entry of packets,
the congestion avoidance mechanism (tail-drop or weighted random early detection
[WRED]) and bandwidth limit must first be checked. When packets leave the queues,
weighted fair scheduling of packets in the queues corresponding to each class should
be performed.
IP Packets
LLQ
BQ1
Classifying
Sent packets
BQ64
Sent queue
¡
¡
Scheduling
BQ2
Outgoing first
Figure 1-1 CBQ diagram
If CBQ performs weighted fair treatment to queues of all classes, voice packets, the
delay-sensitive data flow may not be sent out in time. Therefore, PQ is introduced to
CBQ to create low latency queuing (LLQ), which provides strictly preferred sending
service for such delay-sensitive data flow as voice packets.
LLQ strictly combines PQ with CBQ. When a user defines a class, he can specify it to
accept strict priority service. The class of this type is called priority class. All packets
of the priority class enter the same priority queue. Before they enter a queue, the
bandwidth limit of each class of packets should be checked. When packets go out of
the queues, the packets in the priority queue are forwarded before packets in the
queues corresponding to other classes. But if the maximum reservation bandwidth
configured for LLQ is exceeded, the packets in other queue are sent. Weighted fair
scheduling will be performed to the packets in other queues when they are forwarded.
In order to avoid long time delay of packets in other queues, the maximum available
bandwidth can be specified for each priority class during LLQ application for traffic
3Com Router Configuration Guide Addendum for V1.20
3