A kind of caching method and device

Abstract

The present disclosure discloses a caching method and device, which are applied to a shared cache of network communication equipment. The method includes the following steps: dividing the shared cache into multiple cache areas, and each cache area corresponds to a different priority type; According to the sending priority of the message queue, the message queue is sent to the corresponding buffer area. In the case of more reasonable and efficient use of shared cache resources, this disclosure can ensure queue priority scheduling of different service types, and prevent too many low-priority messages from occupying the cache even when the total cache size is small. High-priority packets enter the queue space to ensure QoS service quality normally and bring users a better experience when surfing the Internet.

Description

technical field [0001] The present disclosure relates to the technical field of network communication, and in particular, to a caching method and device. Background technique [0002] WRED (Weighted Random Early Detection) is a method that monitors the usage of network resources (such as queues or memory buffers) and actively discards packets when congestion tends to intensify. A flow control mechanism to relieve network overload. Its implementation depends on the size of the buffer area. In reality, it is impossible to have an unlimited buffer area size. Once the total buffer area size is exceeded, the WRED mechanism will be abnormal. At the same time, the QoS (Quality of Service) supported by WRED ) queue scheduling will also skew. [0003] The normally forwarded queues basically do not buffer packets, and the traffic is scheduled according to a specific QoS priority, such as figure 1 As shown, each queue will set the maximum buffer usage limit, that is, the total lengt...

AI Technical Summary

Problems solved by technology

However, the disadvantage of the first solution is: cache reservation can only support a small number of queues, cannot solve the problem of abnormal WRED threshold, and cannot effectively solve the problem of QoS scheduling failure

This solution can solve the problem of QoS scheduling failure, but it causes a huge waste of cache usage. In live network applications, the cache is shared by all queues, but the queues are not necessarily all congested. If only half of the queues are congested, half of the queues will be congested. Cache resources will not be used, and the congestion of all queues is rare after all. At the same time, if the total shared cache is not large, the average cache allocated to each queue may not reach the minimum limit of QoS scheduling, that is, evenly allocated After reaching the cache of all queues, the problem of QoS scheduling failure still cannot be solved

Images