End-to-End Link Availability Assessments in an Optical Spectrum as a Service Use-Case

The open and disaggregated approach to optical transport networks has enabled the emergence of new service models such as Optical Spectrum as a Service (OSaaS). In this service model, the Open Line System (OLS) operators provide the spectral resources, allowing end-users to directly access the spectrum and operate their own transceivers over it. In commercial offerings, this service type is typically characterized by link distance, dedicated wavelength range, Optical Signal to Noise Ratio (OSNR), and Generalized OSNR. If the thresholds for the signal power spectral density are met, the OLS operator guarantees the OSaaS availability as agreed in the Service Level Agreement (SLA). When the assigned spectrum can be allocated in multiple consecutive operator domains, the end-to-end services within the OSaaS can traverse over thousands of kilometers, passing multiple network domains, fiber infrastructures, and countries. However, as detailed information about the physical infrastructure and commissioned OLS system is often considered business-critical information, the traversed OLSs appear as black boxes to the end-user. This paper examines the infrastructure, OLS, and transceiver-specific impact on the end-to-end service availability for OSaaS links between 9 and 8000 kilometers. The paper further investigates the accuracy of the link length estimations, using chromatic dispersion (CD) compensation data from 158 live transceivers, operated by four pan-European operators. The analyses show that infrastructure-related failures are dominant already in OSaaS links as short as 37.5 km. For the CD-based link length estimations, an average estimation error of 7.3 % compared to the actual link length was achieved, enabling link length-based availability estimations in OSaaS use-cases operated in black-box scenarios.