A method and device for dispatching messages

Abstract

The invention discloses a method for scheduling a message. The method comprises the steps of determining the length of the maximum message of messages scheduled in the previous round in a first queue, and setting a weight factor of the first queue based on the length of the maximum message, wherein the weight factor is greater than or equal to the length of the maximum message and is less than or equal to a preset threshold value, and the preset threshold value is greater than or equal to the length of the maximum message; calculating a first scheduling length based on the weight factor and a first ratio-item corresponding to the first queue in a weight ratio; and scheduling messages in the corresponding number from the first queue based on the first scheduling length. Therefore, the weight factor of the first queue is dynamically set before each round of the scheduling, at least one message can be scheduled from the first queue in each round of the scheduling, and the problem of message burst caused by the too big weight ratio and the scheduling of excessive messages in one round of the scheduling can be avoided, and the stability of message transmission is accordingly improved.

Description

technical field [0001] The present application relates to the communication field, and in particular to a method and device for scheduling messages. Background technique [0002] The central content of congestion management is how to formulate a strategy when congestion occurs to determine the processing sequence and discarding principle of packet forwarding, generally using queue technology. Queue refers to the logic for ordering packets in a buffer. When the traffic rate exceeds the interface bandwidth or exceeds the bandwidth set for the traffic, the packets are temporarily stored in the cache in the form of queues. The time and sequence of packets leaving the queue and the relationship between packets leaving each queue are determined by the queue scheduling algorithm. [0003] The queue scheduling principle of the weighted differential round robin (English full name: Weighted Deficit Round Robin, English abbreviation: WDRR) scheduling algorithm is as follows: a round ...

AI Technical Summary

Problems solved by technology

In the further improved WDRR algorithm, in order to ensure that the queue can send at least one packet each time it is selected, the service quota is required to be at least the size of the maximum packet length in the queue, otherwise, if the configured service quota is small , when two large packets are dispatched in the last round of scheduling, the overrun service quota is greater than the service quota allocated in the current round of scheduling, then the service quota will be less than zero after deducting the overrun service quota in the previous round, resulting in this Round-robin scheduling fails to schedule packets

[0004] Since the service quota of each queue is at least the size of the maximum message length of the queue, it is guaranteed that at least one message can be sent every time it is selected by scheduling, and the length of the maximum message in the queue is different from other short messages If it is larger, it is possible to send more short messages in one round of scheduling, resulting in serious packet bursts, and the queue may be dispatched after several rounds of scheduling, and subsequent scheduling During the process, only packets from other queues will be dispatched, which will cause excessive traffic jitter during the entire scheduling process and affect the stability of transmitted packets

Images