A method of dynamic load balancing of traffic based on dpdk

Abstract

A kind of flow dynamic load balancing method based on DPDK of the present invention relates to computer multi-core technical field; The method is by setting up two Hash tables; Carry out Hash1 () and Hash2 () calculation respectively, look up table 1 and Hash table 2 according to Hash result, judge Whether there is a mapping relationship between the two Hash calculation results and the processing core id, if yes, distribute the data packet to the corresponding processing core, otherwise, calculate the overall load balance degree of the current CPU; determine whether the load balance degree exceeds the threshold, if not, According to the result of Hash1(), the mapping relationship is established in Hash table 1, and the data packets are allocated to the corresponding processing cores. If yes, the processing cores with serious overload are removed; the mapping between the calculation results of Hash2() and the processing core id is dynamically updated. relationship, according to the result of Hash2 (), the mapping relationship is established in Hash table 2, and the data packets are distributed to the corresponding processing core; the present invention solves the problem of unbalanced multi-core processor load.

Description

technical field [0001] The invention relates to the field of computer multi-core technology, in particular to a DPDK-based flow dynamic load balancing method. Background technique [0002] Today, with the rapid development of information technology, artificial intelligence technology and distributed parallel computing technology, high-performance data processing systems have been widely used in many fields such as telecommunications, finance, industry and medical treatment. The future will be an efficient digital society, where the amount of information, computation, and complexity are rapidly expanding. A large amount of information such as voice, data, and images requires high-speed real-time perception and processing by computers. [0003] Over the years, the improvement of processor performance has mainly relied on the increase of operating frequency. However, due to the influence of factors such as power consumption and heat generation, this approach is close to the lim...

AI Technical Summary

Problems solved by technology

[0007] The disadvantage is that in some network application processing devices, simply using RSS technology will affect the processing performance, such as telecom forwarding equipment, the way of processing a bidirectional stream of a network connection is similar, so it is hoped that the same core of the multi-core processor The data packets with symmetric information are processed on the Internet, and the representative applications include network firewalls and quality of service guarantees.

[0008] Compared with the RSS technology of the DPDK platform, the improved symmetric RSS technology solves the problem that the opposite data packets of the same flow are processed on different cores, but for Hash collisions and large-scale traffic of the same connection, there is still a problem. The load on one core of the multi-core processor is too heavy, and the load is unbalanced

[0009] Therefore, when the network traffic is relatively single and extreme, only using the static hash technology cannot meet the demand for data packet processing, and the single-core load will be overloaded, which will reduce the processing efficiency of the CPU.

Images