H3C Technologies H3C S3600 Series Switches - Manual (Page 741)

1-11
Figure 1-7 Diagram for WFQ queuing
Before WFQ is introduced, you must understand fair queuing (FQ) first. FQ is designed for the purpose
of sharing network resources fairly and optimizing the delays and delay jitters of all the flows. It takes
the interests of all parties into account, such as:
z Different queues are scheduled fairly, so the delay of each flow is balanced globally.
z Both short and long packets are scheduled fairly. When there are multiple long packets and short
packets to be sent among different queues, the short packets must be scheduled preferentially, so
that the delay jitters of packets of each flow is reduced globally.
Compared with FQ, WFQ takes the priority into account when calculating the scheduling sequence of
packets. Statistically speaking, WFQ assigns more scheduling chances to high priority packets than
those to low priority packets. WFQ can classify the traffic automatically according to the session
information of traffic including the protocol types, source and destination TCP or UDP port numbers,
source and destination IP addresses, and priority values in the ToS field. WFQ also provide as many
queues as possible to accommodate each flow evenly. Thus, the delay of each flow is balanced globally.
When the packets dequeue, WFQ assigns the bandwidth to each flow on the egress according to the
traffic precedence or DSCP precedence. The lower the traffic precedence is, the less bandwidth the
traffic gets. The higher the traffic precedence is, the more bandwidth the traffic gets. Finally, each queue
is polled and the corresponding number of packets is taken out to be sent according to the proportion of
bandwidth.
You can use the WFQ algorithm to assign bandwidth to the output queues of a port, and then decide
which queue a traffic flows into according to the mapping between the COS value of the traffic and the
queue, and also deicide how much bandwidth is to be assigned to each traffic.
3) WRR queuing