SRv6-PPO Network Nodes Deployment Method

Aiming at the issues of high economic cost and incomplete consideration of node attributes in the current SRv6 network nodes deployment method, we proposed an SRv6 network node deployment method based on deep reinforcement learning. Firstly, based on constructing a network communication scenario that integrates IPv6 nodes and SR nodes, we proposed an SRv6 network node optimization model to ensure that the selection of SRv6 node locations is more rational. Next, we designed an SRv6 network node optimization model integrating traffic engineering to assess the advantages and disadvantages of SRv6 node deployment locations. Then, an SRv6 network node deployment method based on proximal policy optimization was devised from offline and online phases. Finally, we assessed the proposed method through three typical datasets. The empirical results indicated that, relative to a network environment where all nodes are equipped with SRv6 capabilities, deploying merely 20%-40% of SRv6 nodes within a hybrid network can achieve comparable link utilization. Furthermore, the proposed method significantly outperforms alternative approaches in minimizing the maximum link utilization, thereby enhancing overall network efficiency and stability.