Load sharing method, device and system for message traffic

Abstract

The invention relates to a load sharing method, a load sharing device and a load sharing system for message traffic. The method comprises the following steps of: resolving a received message, and acquiring the stream identification information of a message stream corresponding to the message; querying a message stream state table, and judging whether the message stream is processed by a processor of a multi-core processor, which corresponds to the message stream, or not according to the stream identification information; if the message stream is processed by the processor, arranging the message into the message stream, and processing the message by using the processor; if the message stream is not processed by the processor, arranging the message into the message stream, and distributing the message stream to an idle process of the multi-core processor for processing; and judging whether the processing of the message stream is finished or not, and if the processing of the message stream is finished, updating the message stream state table for the distribution of the next message stream. The problem of conflict between traffic load sharing and message sequence preservation in a multi-core system is solved, the dynamic balanced sharing of the traffic in the multi-core system is realized, the overhead of the message sequence preservation is eliminated, and message forwarding performance improved.

Description

technical field [0001] The present invention relates to the field of communication, in particular to a method, device and system for load sharing of message flow. Background technique [0002] In the field of communication, there are a large number of high-speed message forwarding and processing scenarios. With the processor CPU, ASIC (Application Specific Integrated Circuit), field-programmable gate array With the continuous development of technology, the high-speed packet forwarding technology is also constantly evolving. Due to the strong computing power of general-purpose CPUs, it is not suitable for forwarding processing of large-bandwidth I / O access; ASIC / FPGA has strong forwarding capabilities but insufficient flexibility; although network processors combine the high performance of ASICs and the programmable capabilities of CPUs, However, it cannot adapt to the rapid and diversified evolution of wired or wireless services. The multi-core processors that have emerged...

AI Technical Summary

Problems solved by technology

Packet distribution is sequentially distributed to multiple cores for processing one by one. Each concurrent thread covers all business processes and executes the same code concurrently. The synchronization operation of related messages is completed through the mutual exclusion mechanism. In this way, although the multi-core CPU utilization is relatively high However, the Cache hit rate is reduced, and the mutex has a greater impact on performance

The flow distribution method is based on the data flow method and is distributed to multiple cores for processing. Although the introduction of mutexes is avoided, there is a problem of unbalanced multi-core load when single-flow traffic surges.

Some CPUs are overloaded, some are too light, and the advantages of multi-core load sharing are not fully utilized

Images