CN-122027498-A - Packet transmission optimization method and router based on intelligent traffic prediction

CN122027498ACN 122027498 ACN122027498 ACN 122027498ACN-122027498-A

Abstract

The invention relates to the technical field of routers, in particular to a packet transmission optimization method based on intelligent flow prediction, which comprises the following steps of S1, extracting flow characteristics, S2, generating a flow prediction result, S3, dynamically generating a packet transmission optimization strategy aiming at a preset time window in the future, and S4, actively optimizing packet transmission. The intelligent flow prediction-based packet transmission optimization method realizes the fundamental change of network flow management, namely, the network flow management is upgraded from a traditional passive and reactive management mode to a prediction-based active and preventive optimization mode, and a router can pre-configure resources and adjustment strategies before flow congestion or service quality reduction occurs through a closed loop of acquisition-prediction-strategy generation-execution, so that network jitter is obviously reduced, transmission delay is reduced, sudden packet loss is avoided, and smoother, more reliable and predictable network experience is provided for users.

Inventors

ZHOU NENGQIANG

Assignees

安徽数航科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (8)

1. The packet transmission optimization method based on intelligent flow prediction is characterized by comprising the following steps: s1, acquiring and analyzing historical flow data in a network in real time, and extracting flow characteristics; s2, based on the extracted flow characteristics, predicting a network flow mode in a future preset time window by utilizing a pre-trained flow prediction model to generate a flow prediction result, wherein the flow prediction result at least comprises predicted flow size, application type distribution and corresponding service quality requirements; s3, dynamically generating a packet transmission optimization strategy aiming at the preset time window in the future according to the flow prediction result, wherein the optimization strategy comprises at least one of a dynamic service quality strategy and a path selection strategy; And S4, when the preset time window arrives, the packet transmission optimization strategy is applied to a data plane of the network equipment, and active optimization is carried out on packet transmission.
2. The packet transmission optimization method based on intelligent flow prediction according to claim 1, wherein in the step S2, the pre-trained flow prediction model is a hybrid prediction model and comprises a cyclic neural network model for short-term prediction and a time sequence decomposition model for long-term pattern recognition, and the step S2 is characterized in that the step S is used for generating a flow prediction result specifically comprises the steps of calling the cyclic neural network model to conduct minute-level flow prediction, and fusing periodicity and trend characteristics output by the time sequence decomposition model to generate a comprehensive prediction result.
3. The method for optimizing packet transmission based on intelligent traffic prediction according to claim 1, wherein step S2 further comprises performing online learning and parameter fine tuning on the traffic prediction model according to the difference between the traffic data collected in real time and the prediction result, so as to adapt to the change of the network traffic mode.
4. The packet transmission optimization method based on intelligent traffic prediction according to claim 1, wherein in step S3, the dynamically generated packet transmission optimization policy specifically includes: S31, dynamically distributing priority and bandwidth guarantee parameters for data streams of different application types based on application type distribution in the flow prediction result to form a dynamic service quality strategy; s32, calculating optimal forwarding paths for data flows with different priorities and destinations based on the flow size and distribution of the flow prediction result and combining network topology and link states to form a path selection strategy.
5. The packet transmission optimization method based on intelligent traffic prediction as recited in claim 1, wherein the optimization strategy generated in step S3 further comprises a cache prefetch strategy, and the method further comprises prefetching part of the content data in advance in a traffic idle period or a network edge node according to the prediction of the video stream or the content download stream.
6. The packet transmission optimization method based on intelligent traffic prediction according to claim 1, wherein in step S4, the packet transmission optimization policy is applied to the data plane, specifically, the policy is issued to a forwarding device supporting programmable packet processing through a software defined network controller, and the forwarding device marks, queues, and forwards the packet according to policy rules.
7. A packet transmission optimized router based on intelligent traffic prediction according to claims 1-6, characterized by comprising: the flow collection and analysis module is used for collecting and analyzing historical flow data in the network in real time and extracting flow characteristics; the flow prediction module is connected with the flow acquisition and analysis module and is used for predicting a network flow mode in a future preset time window by utilizing a pre-trained flow prediction model based on the extracted flow characteristics to generate a flow prediction result; the strategy generation module is connected with the flow prediction module and is used for dynamically generating a packet transmission optimization strategy for the preset time window in the future according to the flow prediction result; the policy execution module is connected with the policy generation module and is used for executing the packet transmission optimization policy when the preset time window arrives, and actively optimizing the packet transmission through the router; The flow prediction result at least comprises predicted flow size, application type distribution and corresponding service quality requirements, and the optimization strategy comprises at least one of a dynamic service quality strategy and a path selection strategy.
8. The router for packet transmission optimization based on intelligent traffic prediction according to claim 7, further comprising a hardware acceleration module, wherein the hardware acceleration module comprises a neural processing unit and a programmable switching chip, the neural processing unit is dedicated to running the traffic prediction model, and the programmable switching chip is used for executing the packet processing and forwarding rules issued by the policy execution module at a high speed.

Description

Packet transmission optimization method and router based on intelligent traffic prediction Technical Field The invention relates to the technical field of routers, in particular to a packet transmission optimization method based on intelligent traffic prediction and a router. Background With the explosive growth of applications such as the internet of things, high definition video streaming, online collaboration and cloud gaming, modern networks, particularly home and small and medium-sized enterprise networks, are facing increasingly complex and dynamically changing traffic loads. Conventional routers act as core scheduling devices for network traffic and their modes of operation are passive in nature. They rely primarily on static routing tables, simple quality of service rules (e.g., fixed priority based on port or protocol), and reactive scheduling algorithms such as "first come first served" or weighted fair queuing to manage packet forwarding. The traditional method only carries out processing decision when the data packet arrives, cannot predict the flow change in the short-term future, and when the sudden high-priority flow and the background high flow occur simultaneously, the network is extremely easy to generate congestion, so that key application is blocked, delay is increased, and user experience is reduced sharply. In order to solve the problems, an improvement is provided, and a packet transmission optimization method and a router based on intelligent traffic prediction are provided. Disclosure of Invention In order to solve the technical problems, the invention provides the following technical scheme: the invention provides a packet transmission optimization method and a router based on intelligent flow prediction, comprising the following steps: s1, acquiring and analyzing historical flow data in a network in real time, and extracting flow characteristics; s2, based on the extracted flow characteristics, predicting a network flow mode in a future preset time window by utilizing a pre-trained flow prediction model to generate a flow prediction result, wherein the flow prediction result at least comprises predicted flow size, application type distribution and corresponding service quality requirements; s3, dynamically generating a packet transmission optimization strategy aiming at the preset time window in the future according to the flow prediction result, wherein the optimization strategy comprises at least one of a dynamic service quality strategy and a path selection strategy; And S4, when the preset time window arrives, the packet transmission optimization strategy is applied to a data plane of the network equipment, and active optimization is carried out on packet transmission. In step S2, the pre-trained flow prediction model is a mixed prediction model and comprises a cyclic neural network model for short-term prediction and a time sequence decomposition model for long-term pattern recognition, and the generation of the flow prediction result specifically comprises the steps of calling the cyclic neural network model for minute-level flow prediction, and fusing the periodicity and the trend characteristics output by the time sequence decomposition model to generate a comprehensive prediction result. As a preferable technical scheme of the invention, according to the difference between the flow data collected in real time and the prediction result, the flow prediction model is subjected to online learning and parameter fine adjustment so as to adapt to the change of the network flow mode. As a preferred technical solution of the present invention, in step S3, the dynamically generated packet transmission optimization policy specifically includes: S31, dynamically distributing priority and bandwidth guarantee parameters for data streams of different application types based on application type distribution in the flow prediction result to form a dynamic service quality strategy; s32, calculating optimal forwarding paths for data flows with different priorities and destinations based on the flow size and distribution of the flow prediction result and combining network topology and link states to form a path selection strategy. The optimization strategy generated in the step S3 further comprises a cache prefetching strategy, and the method further comprises prefetching part of content data in advance in a traffic idle period or a network edge node according to prediction of video streams or content download streams. In step S4, the packet transmission optimization strategy is applied to the data plane, specifically, the strategy is issued to the forwarding device supporting the programmable data packet processing through the software defined network controller, and the forwarding device marks the data packet, queues the scheduling and forwards the path according to the strategy rule. A packet transmission optimized router based on intelligent traffic prediction, comprising: the flow colle