.png){kind=icon}
.png){kind=icon}
Return
Federated Learning enabled software-defined optical network with intelligent control plane architecture
.png){kind=icon}
PRE
AI
DOI：10.1016/j.compeleceng.2024.109329.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Software-defined optical network (SDON) is a software-defined networking (SDN) architecture that incorporates an optical network in the data plane. By integrating machine and deep learning (ML/DL) techniques into SDON, the network can make intelligent decisions and enhance its performance. However, these intelligent architectures face various challenges, such as increased burst loss and propagation delay due to selecting the same model at each node for prediction, a huge centralized database, and higher response latency between layers. Therefore, a modified federated learning (FL) SDON architecture is proposed. In the proposed architecture, data collection is performed at each terminal node to avoid centralized storage of data. Due to the varying and evolving nature of the network data patterns, different ML and DL models, according to the requirements, are implemented on different nodes, to predict burst loss and propagation delay. To reduce the mean response latency, models are trained on the terminal node and as a result, the least communication happens between data and the control plane. Based on the dataset created using data collected from a network representing real-world conditions, we observe that the proposed architecture outperforms the existing state-of-the-art by accurately predicting 98.18% times burst loss and 99.85% times propagation delay.
Keywords:
Software-defined network
Optical network
Federated learning
Burst loss
Propagation delay
Response latency
Journal
C
IF:
4.9
Papers:
6.7K
Citations:
1.3W