Research Article
Deep Reinforcement Learning Based Congestion Control Mechanism for SDN and NDN in Satellite Networks
@INPROCEEDINGS{10.1007/978-3-031-34497-8_2, author={Ziyang Xing and Hui Qi and Xiaoqiang Di and Jinyao Liu and Ligang Cong}, title={Deep Reinforcement Learning Based Congestion Control Mechanism for SDN and NDN in Satellite Networks}, proceedings={Mobile Wireless Middleware, Operating Systems and Applications. 11th EAI International Conference, MOBILWARE 2022, Virtual Event, December 28-29, 2022, Proceedings}, proceedings_a={MOBILWARE}, year={2023}, month={5}, keywords={satellite network congestion control deep reinforcement learning ICN}, doi={10.1007/978-3-031-34497-8_2} }- Ziyang Xing
Hui Qi
Xiaoqiang Di
Jinyao Liu
Ligang Cong
Year: 2023
Deep Reinforcement Learning Based Congestion Control Mechanism for SDN and NDN in Satellite Networks
MOBILWARE
Springer
DOI: 10.1007/978-3-031-34497-8_2
Abstract
In a satellite network, the content-centric information center network can reduce redundant data and decouple the location of network entities from the content, which is especially suitable for sending massive data from satellites to the ground. Influenced by outages, the information center network congestion control is inefficient and adaptive, the congestion policy cannot be changed from a global perspective, and the paths saved in the FIB and PIT are prone to failure. This paper proposes a congestion control algorithm based on deep reinforcement learning: RL-ICN-CC, which uses a software-defined network controller to obtain the state information of the whole network, deep reinforcement learning realizes an adaptive congestion control mechanism, and consumers adjust the interest packets according to the global state of the network CWND of the sending rate to avoid congestion. In this paper, FIB and PIT are redesigned. When the saved path easily fails, consumers can still calculate other cache locations to obtain content. Compared with other algorithms in multiple scenarios, the throughput of the proposed scheme is improved by 11%, and congestion adaptability is achieved.
