In a bold step towards revolutionizing network technologies, a Japanese/German research collaboration has launched a platform assisting prophylaxis against breakdowns in optical networks. The new testbed supports the development and testing of AI and Machine Learning models. Notably, the solution enables the use of telemetry data for network failures. This type of data would often be classified and therefore inaccessible to researchers.
The platform, the Optical Testbed Dataspace (OTDS), is a joint initiative by the National Institute of Information and Communications Technology (NICT), Japan, and the Fraunhofer Heinrich-Hertz Institute (HHI), Germany.
“The data sovereignty framework, such as OTDS, is essential to enable open innovation in AI-driven networks according to today’s global context. OTDS would also further enhance the importance of the network testbeds operated by the respective institutions,” says Dr. Awaji Yoshinari, Director General of the Photonic ICT Research Center at NICT.
AI training requires big data
Dataset availability is a significant challenge in advancing networks with AI and Machine Learning research. Network data is typically classified by operators and regulatory bodies, while vendors restrict access to specific equipment telemetry. Alternatives, such as experimental or synthetic data, often result in overfitting and poor generalization performance.
To address this, optical network testbeds are evolving to play a crucial role in delivering big data – specifically, telemetry data for rare and abnormal events, such as network failures. This data is essential for training and validating AI-assisted network functions.
“Our mission is to design state-of-the-art innovations without compromising on data privacy or safety. OTDS enables the secure exchange of network data, fostering the development of innovative AI models that comply with strict data privacy and export control regulations,” says Dr. Johannes Fischer, Head of the Digital Signal Processing Group at Fraunhofer HHI.
Lowering the barrier for data access
OTDS provides a framework for testbeds to securely share their data while ensuring compliance with data sovereignty and export control requirements. This allows research institutions, network operators, and vendors to exchange valuable data with defined access and usage controls, without sacrificing control or breaking regulatory guidelines.
“Real-time data exchange from testbeds generates diverse datasets that are crucial for training and validating AI models, facilitating the automation and optimization of network functions. OTDS lowers the barriers to data access and fosters an open research environment, paving the way for the rapid validation of cutting-edge network automation solutions,” explains Dr. Behnam Shariati, Head of the “AI for Photonics” division at Fraunhofer HHI.
“With OTDS, we can strengthen global research collaboration and accelerate the development of the next generation of AI-powered optical networks, improving mobile connectivity for remote areas and network resiliency, or support real-time data processing in industry and critical infrastructures,” says Dr. Yoshida Yuki, Research Manager of the Photonic ITC Research Center at NICT.
All details on the new testbed are published in the paper “International Testbed Data Sharing Framework with Data Sovereign Features for Network AI/ML Empowerment.”
Source: “NICT and Fraunhofer unveil groundbreaking international optical testbed data space to enable AI/ML-driven networks”, joint press release from the two institutions.
Contributors
