Multi-core message forwarding method, multi-core processor and network equipment

Abstract

The invention provides a multi-core message forwarding method, a multi-core processor and network equipment. The method comprises the following steps of: storing received messages into message groups corresponding to different improvement pipelines in a hash way so that the messages of the same message flow are stored in the message group corresponding to the same improvement pipeline in the hash way; acquiring virtual thread objects from a global virtual thread object sequence by using a processing core in the multi-core processor according to the order of the priority levels from high to low, wherein the global virtual thread object sequence stores the virtual thread objects in a ready state at each priority level, and the virtual thread objects in the ready state at the same priority level are stored in the global virtual thread object sequence according to a first-in first-out order; and executing a processing function in the acquired virtual thread object to finish forwarding a message to be processed corresponding to a private object pointer in the acquired virtual thread objects by using the processing core. By using the technical scheme provided by the invention, the parallel processing capacity of the multi-core processor is improved.

Description

technical field [0001] The invention relates to communication technology, in particular to a multi-core message forwarding method, a multi-core processor and network equipment. Background technique [0002] With the deepening of social informatization, people's dependence on the Internet is getting stronger and stronger, and the number of computer nodes connected to the network is also increasing exponentially. Network devices, etc. are used as data exchange nodes for computer interconnection, and their data forwarding performance determines the quality of the entire network. With the rapid increase in the number of user connections and the emergence of various high-bandwidth applications, it is increasingly difficult for single-core processors with limited main frequency to meet the performance requirements of high-speed data forwarding. The characteristics of scalability and parallel computing have gradually been widely used in the field of network communication. The so-...

AI Technical Summary

Problems solved by technology

[0003] However, since multiple processing cores that execute message forwarding processing programs in parallel share many hardware and software resources, when data forwarding is busy, how to make each processing core cooperate to participate in forwarding, reduce mutual waiting between each other, and give full play to parallel processing Ability is very challenging

In addition, due to the randomness of the program running by the processor, the time for each processing core to process the message is not necessarily the same, resulting in the message that enters first may be output later, resulting in the problem of out-of-order messages. For a message flow, Out of order after being forwarded by network equipment is not allowed

[0004] In order to take advantage of the parallel processing capabilities of multi-core processors and solve the problem of out-of-sequence messages in the above-mentioned message forwarding process, the following solutions exist in the prior art: pipeline mode, which divides the entire message forwarding process into multiple processing stages (Stage ), each processing core is responsible for a processing stage, forming a serial message processing pipeline, but the characteristic of this method is that each processing stage of the pipeline is bound to a fixed processing core, and each processing core will only participate in a unique In one processing stage of a pipeline, if the division of tasks in each pipeline or the same pipeline is unbalanced, some processing cores may be very busy, even exceeding their processing capacity, while other processing cores are very idle, but because each processing The core can only participate in the processing stage bound to itself, so it cannot help the busy processing core, which limits the improvement of the parallel capability of the multi-core processor

Packet flow and processing core binding mode, that is, using a pre-packet flow hash unit to distribute packets of different packet flows to different processing cores, ensuring that the same packet flow is distributed to a certain processing core, while The entire forwarding process of the message flow is completed by the processing core. However, the defect of this scheme is that the processing burden of different message flows may vary greatly. Even if the message flow hashing unit implements balanced hashing, some Processing cores are relatively idle, while other processing cores are relatively busy, which also limits the improvement of the parallel capability of multi-core processors

Images