Message scheduling method and device and network chip

Abstract

The invention provides a message scheduling method and device and a network chip, which are applied to the network chip. The network chip comprises a classification layer, and the classification layercomprises a plurality of classification nodes; the method comprises the following steps: after a classification node acquires a scheduling task of a network message, and acquiring a node identifier of the classification node; determining a queue identifier corresponding to the node identifier of the classification node according to a mapping relationship between the stored node identifier and thequeue identifier of the output queue; judging whether an output queue corresponding to the determined queue identifier meets a back pressure condition or not; when a back pressure condition is satisfied, giving up the scheduling task; when the back pressure condition is not met, executing a scheduling task of the network message, so that the network message is written into the determined output queue. By the adoption of the method, scheduling of the network messages is achieved on the L3 layer, the congestion problem caused by the fact that the cache amount of the output queues is too large is effectively avoided, and the cache pressure of the corresponding output queues is reduced.

Description

technical field [0001] The present application relates to the technical field of computer communication, in particular to a message scheduling method, device and network chip. Background technique [0002] With the explosive growth of traffic in the network, a good traffic management method is particularly important. In the network chip, traffic congestion management and congestion avoidance are mainly implemented in the traffic management TM unit. The TM unit internally schedules the received messages. The internal scheduling topology of the TM is divided into five layers L4~L0. The layers from L4 to L0 are flow layer, class layer, subport layer, port layer, and interface layer. The layer performs strict scheduling processing on the message, and sends the scheduled message to the output queue (Output Queue, OQ). However, the OQ of the output queue has a capacity limit. Once the output queue traffic buffer reaches the queue threshold, it is necessary to perform back pressu...

AI Technical Summary

Problems solved by technology

[0003] The current backpressure mechanism is based on interface level or port level backpressure, that is, once the number of interface layer / port layer packets is found to exceed the threshold, backpressure will be applied to L0 layer or L1 layer to block the flow through L0 layer or L1 layer message traffic, but this solution will prevent all traffic of L0 layer or L1 layer nodes that are currently under pressure from being sent externally, which will cause some important traffic to be unable to be sent preferentially

Images