Weighted fair queueing (WFQ) is a method of automatically smoothing out the flow of data in packet-switched communication networks by sorting packets to minimize the average latency and prevent exaggerated discrepancies between the transmission efficiency afforded to narrowband versus broadband signals. In WFQ, the priority given to network traffic is inversely proportional to the signal bandwidth
. Thus, narrowband signals are passed along first, and broadband signals are buffered.
WFQ has little or no effect on the speed at which narrowband signals are transmitted, but tends to slow down the transmission of broadband signals, especially during times of peak network traffic. Broadband signals share the resources that remain after low-bandwidth signals have been transmitted. The resource sharing is done according to assigned weights. In flow-based WFQ, also called standard WFQ, packets are classified into flows according to one of four criteria: the source Internet Protocol address (IP address), the destination IP address, the source Transmission Control Protocol (TCP) or User Datagram Protocol (UDP) port, or the destination TCP or UDP port. Each flow receives an equal allocation of network bandwidth, hence the term fair.
There are two other forms of WFQ, known as VIP-distributed WFQ for VIP2-40 or greater interface processors, and class-based WFQ in which the the traffic is categorized into user-defined classes. Both of these forms of WFQ operate according to principles similar to that of standard (flow-based) WFQ.
WFQ can prevent high-bandwidth traffic from overwhelming the resources of a network, a phenomenon which can cause partial or complete failure of low-bandwidth communications during periods of high traffic in poorly managed networks.