Service flow information-based link state prediction method for an SDN

Abstract

The invention discloses a service flow information-based link state prediction method for SDN. The method comprises the following steps of acquiring data flow in a link, counting the data flow according to a protocol to obtain a statistical index, obtaining a calculation index according to the statistical index, and predicting the link state according to thecalculation index. According to the method, the statistical index is obtained by counting the data flow according to the protocol, the calculation index containing the service information is obtained through the statistical index, and the calculation index is used as the input of the prediction model for final prediction. As the calculation index contains the service category information (such as the data volume transmitted by each service and the proportion of the number of different services), the input granularity is updated, and the accuracy of link state prediction is effectively improved.

Description

technical field [0001] The invention belongs to the technical field of SDN network, and in particular relates to a link state prediction method and system of an SDN network based on service flow information. Background technique [0002] A software-defined network (Sofeware Defined Network, SDN) has a globally controllable control plane and an efficient forwarding plane, and decision makers can schedule and optimize network resources based on global information. Global information includes the network topology and the link status of each link in the topology (such as the used link bandwidth, delay, packet loss rate, etc. of the link). To achieve accurate scheduling, SDN needs to perceive links in a timely and accurate manner. state. [0003] The current link state perception methods mainly include measurement and prediction. The former uses active measurement or passive measurement to obtain the link status of the current network; the latter analyzes the change trend of tr...

AI Technical Summary

Problems solved by technology

However, because the current service flow information measurement method can only obtain service flow information at the end nodes through active packet sending or plug-in monitoring, it is difficult for the intermediate nodes of the network to passively obtain upper-layer application information. The existing neural network-based link state prediction methods only Rely on information below the application layer for prediction, the input is still at a coarse granularity, and the prediction accuracy is not high

Images