How do we learn from optical networks to better automate them?
Read part one of a blog series that explores how optical network operators can use automation to unlock cost savings and speed up optical service delivery.
The complexities of networking light can make running an optical network a costly and challenging endeavor. Many optical network operators would like to reduce network total cost of ownership (TCO) by automating their optical networks. But the stakes are high: an error can obliterate petabits of valuable data.
That said, light contains a wealth of information that can help us automate the tuning of optical wavelengths to optimize network capacity. To learn from the transmission of light, we need:
- Fully instrumented optical equipment with integral coherent digital signal processors (DSPs) such as the Nokia PSE, which can generate precise telemetry when transmitting and receiving light
- The ability to securely capture and store large amounts of data
- Optical network data science experts who can understand and analyze data to extract useful indicators
- Physics experts who can translate indicators into physical measurements to help network designers and operators take fast, smart and reliable actions in the network
- Expertise for hardening measurements to make them as reliable as measurement standards
- Tools to support the visualization of data and the application of techniques such as machine learning to automate tasks and predict events
- The ability to embed the knowledge gained above into applications that are easy to deploy and use
With the resulting applications in place, network operators have a foundation for gaining a deeper understanding of how optical networks work. This knowledge can make it easier to automate optical networks. It can also enable operators to meet new service demands by increasing network capacity using algorithms that intelligently balance wavelength signal strength, capacity and availability.
This post is the first in a series that looks at what we can learn from optical networks and how we can use this knowledge to support automation that reduces network TCO and accelerates service delivery. The series will feature posts from Nokia Bell Labs optical networks, data science and physics experts who are investigating new ways to extract knowledge and value from the network.
These experts will share what they have learned about:
- How data science techniques such as machine learning can be used to automate the tuning of optical networks and create a more agile service-supporting infrastructure as companies unite IT systems with networks to make them more responsive to business needs
- How to reduce the risk associated with optical network automation by using multiple approaches to fine tune wavelength power and optimize the trade-off between network capacity and availability
- How to apply data science techniques to wavelength performance indicators to accurately predict network fiber failures and initiate proactive protection and restoration measures that minimize their impact on services
This research is being validated in live customer network tests and it’s showing promising results. Algorithms designed to predict outages and take corrective action against them would have successfully prevented past outages. Algorithms for safely adjusting wavelength modulation rates for optimal network performance are also paving the way towards more economical approaches for running optical networks. Several of these algorithms have been endorsed by the scientific community and documented in prestigious scientific journals.
These experts are also actively working with our R&D groups to enhance the Nokia WaveSuite Network Insight Health & Analytics application. We’re building their knowledge and findings into easy-to-use tools that will help network operators maximize the potential of their networks.
Before we get to this exciting research, we need to discuss what it takes to capture quality data that can enable us to automate optical networks in new ways to maximize performance and reduce network TCO. The next post in the series will explore why streaming telemetry and data modeling play an important role in facilitating the efficient capturing of data that can be used to automate optical networks.
Learn more about automating optical networks