Network traffic scheduling method and system based on deep reinforcement learning

Abstract

The invention discloses a network traffic scheduling method based on deep reinforcement learning, and aims to provide intelligent routing and explore, utilize and learn an optimal path when a routing decision is made by utilizing network state information. According to the method, all link state indexes in the whole network are calculated by collecting network information, a reward function related to bandwidth, time delay and packet loss rate is designed, path state information is explored and learned by adopting a double deep Q learning network DDQN, and an optimal path between each pair of source and destination nodes is obtained. According to the method, an optimal route is installed on a switch of a data layer in advance by utilizing a network global view provided by an SDN and interaction between a DRL intelligent agent and an environment; the invention also provides a traffic scheduling system based on the method, which adopts a four-layer SDN structure comprising a knowledge layer, a management layer, a control layer and a data layer, and realizes dynamic scheduling of network traffic along with an actual environment.

Description

technical field [0001] The invention relates to the field of network technology, in particular to a network traffic scheduling method and system based on deep reinforcement learning. Background technique [0002] With the rapid development of the Internet, people have higher requirements for WAN data transmission. On the one hand, greater bandwidth guarantee is required, and on the other hand, the reliability of data transmission has been raised. In this case, how to select links that meet the QoS requirements has become a current research hotspot. [0003] Based on the earlier statement that "innovation to promote the network requires programming on the hardware data path", that is, the dynamic network, the researchers' initial idea was to layer and separate data from control. With the research going on, the three-layer SDN (Software Defined Network) came into being. The decoupling of the underlying data is realized, and the data is separated from the control layer. Tra...

AI Technical Summary

Problems solved by technology

Traditional research ideas use OSPF and other traditional algorithms for link selection in the control layer of the SDN architecture, but there are great limitations. The traditional algorithm only selects the link with the shortest number of hops, not the link with the smallest delay or the largest bandwidth. road, it is easy to cause link congestion

On this basis, Q-learning provides a good help for pathfinding. Q-learning based on Markov decision traverses all possible actions for each state, so as to select the optimal action. The limitation is that when the state and action When gradually increasing, the traversal efficiency becomes low and a lot of storage space is wasted

Images