Nokia and Orange France have tested a way of diagnosing fibre-network problems using data already generated by optical transmission equipment, potentially giving operators deeper visibility into infrastructure without installing another layer of monitoring hardware.

The live trial ran on Orange France’s optical transport network and processed thousands of data points from coherent transmissions. Algorithms developed by Nokia Bell Labs used those measurements to infer conditions along individual links, identifying problems including fibre degradation, amplifier faults, and configuration errors.

Network tomography applies the logic of medical imaging to communications infrastructure: rather than directly instrumenting every component, software uses observations collected across the network to infer the condition of sections that are otherwise difficult to inspect continuously.

That distinction can carry significant operating value because national fibre networks are geographically dispersed and expensive to investigate manually. A service problem may originate in fibre condition, amplification, configuration, or equipment elsewhere on the route, while the customer experiencing degraded performance gives the operator little indication of where the underlying fault sits.

Nokia says the trial produced near-real-time visibility across Orange’s optical infrastructure and allowed degradation to be identified before customer service-level agreements were affected. Because the approach uses equipment already deployed in the network, operators could gain additional observability without adding dedicated probes throughout every route.

Christian Gacon, vice-president of broadband networks at Orange France, said: “By using smart automation technology, we can better optimize the network and deliver more resilient services to our customers.”

The trial arrives while telecom networks are absorbing a change in both the volume and character of traffic. Video remains substantial, but cloud services, distributed applications, interconnected data centres, and AI workloads increasingly depend on predictable high-capacity links between locations that can be hundreds or thousands of kilometres apart.

Optical transport therefore sits underneath much of the AI infrastructure investment currently concentrated on chips and data centres. Accelerators in separate computing campuses create little economic value if data cannot move between them, storage systems, cloud regions, and users with sufficient capacity and reliability.

Automation needs better network visibility

Telecom operators have spent years automating networks that were historically managed through separate hardware systems and highly specialised engineering teams. Greater automation can reduce manual intervention, but software can only make useful decisions when the information describing network condition is sufficiently accurate.

Tomography could add another source of operational evidence by turning ordinary transmission behaviour into estimates of physical network health. If those estimates remain reliable across larger deployments, operators could detect gradual deterioration earlier, investigate commissioning problems more quickly, or direct engineers towards sections of infrastructure where intervention is actually needed.

Capacity utilisation is another potential benefit. Optical networks are normally engineered with safety margins because link conditions vary and operators cannot assume theoretical performance under every circumstance. Better information about actual conditions can allow a network to be operated closer to its usable limits without simply discarding resilience.

Nokia says the trial showed improved potential for spectral efficiency, which determines how effectively available optical spectrum is used. Small gains can become commercially significant across large national networks because extracting more capacity from installed fibre and equipment is generally cheaper than building another route.

The longer-term objective is closed-loop operation, where monitoring identifies a problem and software adjusts network configuration or capacity with limited manual involvement. Telecom vendors frequently describe that destination as autonomous networking, although production operators remain cautious about giving software unrestricted control over infrastructure whose failures can affect large numbers of customers.

Network tomography is consequently more credible as an input to automation than as evidence that optical networks can now operate themselves. Better diagnostics can reduce uncertainty while operators continue deciding which remedial actions are safe to automate and which still need engineering approval.

There is a potential environmental benefit as well. Avoiding unnecessary field interventions and making better use of installed equipment could reduce travel and postpone some infrastructure expansion, although the scale of those savings would need to be demonstrated across production networks rather than inferred from a single trial.

Mixed infrastructure will provide another challenge. Large operators rarely run one generation of equipment from one supplier across every route, which means commercial adoption will depend on whether tomography works consistently across older assets, different fibre conditions, and operational systems built up over many years.

The Orange deployment still represents a meaningful step because it took the technology onto a live service-provider network rather than confining it to a laboratory. That exposes the algorithms to the noise, variation, and incomplete information found in real infrastructure.

As AI investment increases the value placed on high-capacity connectivity, operators will need to extract more performance from networks they already own while identifying degradation before it becomes visible to customers. Reading the condition of fibre through the signals already crossing it offers one route towards both objectives without instrumenting the network again from scratch.