Build better MOFNs using network automation to power hyperscaler growth
Managed optical fiber networks, or MOFNs, describe an arrangement between communications service providers (CSPs) and hyperscalers that provides a middle ground between traditional managed-bandwidth services and private optical builds. While the MOFN term has only been in use for a few years, the underlying concepts have been under development for decades.
Driving the demand for MOFNs is hyperscalers’ insatiable need for bandwidth, a situation compounded by AI traffic. A recent industry survey, for example, found CSPs believe AI traffic will soon account for between 30% and 50% of long-haul traffic by 2027. As hyperscalers invest tens of billions of dollars each year on new datacenters to process AI workloads, the optical networks connecting them have become a bottleneck – that same survey found only 16% of CSPs feel their network is “very ready” for AI traffic today.
This urgent need for bandwidth demanded new approaches to infrastructure planning as network-buildout requests are adopted by CSPs. Hyperscalers face regulatory restrictions that prevent them from building or operating transport optical networks in many jurisdictions. They may also face shortages of the required optical-engineering expertise. In addition, they encounter difficulties laying new fiber quickly enough to reduce time-to-market.
The emergence of MOFNs addressed the shortcomings of both managed bandwidth and private networks by creating a shared-responsibility model. CSPs that operate in the regions where hyperscalers have data centers could either leverage existing infrastructure or expand it to provide the necessary connectivity—without hyperscalers having to assume the full risks and operational burdens of managing a network themselves. It also included the ability for hyperscalers to own their transponders and connect directly into the CSP’s open optical line systems. This meant hyperscalers could influence the optical design across data center locations, specify the required performance metrics and still avoid the operational overhead of managing a transport network.
While the deployment of MOFNs has progressed, they have not grown at the rate CSPs expected because of the slow pace of provisioning infrastructure, deploying wavelength services and obtaining the necessary assurance insights to ensure SLA compliance. In other words, even after the network is built, the time to market is still too long.
By upgrading the MOFN formula to include a visibility layer, spectrum services and multivendor management, however, the future still looks bright for this approach. Nokia recently published a white paper on why the MOFN model will be an essential part of the next phase of cloud and AI infrastructure delivering continued benefits to CSPs and hyperscalers.
Here are a couple of the key findings:
- How can a new MOFN operational model better serve the needs of hyperscalers and CSPs?
One trade-off with early MOFN deployments is a lack of network resource visibility. CSPs operate the entire network, so hyperscalers can’t readily access information on service health, alarms, performance metrics or potential SLA issues. Instead, they must rely on the CSP’s assurance processes, which creates uncertainty and slows troubleshooting. In addition, hyperscalers business models require detailed assurance metrics to ensure they adhere to strict SLAs with their end customers.
Nokia has proposed enhancements to MOFN to solve this issue. It consists of a hierarchical network resource domain insight and monitoring layer that gives hyperscalers read-only access to the physical and logical network topology, operational state of the network and key performance indicators for their portion of the network.
The result is a “smart” MOFN model where CSPs maintain control over operations, but hyperscalers get the transparency they need to treat the network as an extension of their cloud infrastructure. This visibility becomes even more valuable when hyperscalers partner with multiple CSPs in a region. Nokia’s centralized management platform provides a unified view of all MOFN deployments — even across regional, interconnected split networks.
- How can MOFN support emerging spectrum services and multivendor, multi-network interconnect?
AI-scale data center interconnect can require terabits of capacity between sites. While dark fiber may be available from CSPs, laying out equipment between data center locations takes time, and in some regions dark fiber is scarce and building new fiber routes takes even longer. Spectrum sharing provides an alternative in which hyperscalers obtain slices of optical spectrum within the C-band or L-band on an existing shared line system using their own coherent optics before considering a new build over an entirely new dark fiber. CSPs benefit because they can monetize unused spectral resources and reduce bottlenecks created by limited fiber availability.
Nokia’s MOFN framework supports this spectrum service model by leveraging automation along with the hierarchical management layer to support it. CSPs can publish the spectrum they have available, and hyperscalers can request, modify or remove connections through an automated fulfillment process. This ensures spectrum slices are reserved, provisioned and monitored efficiently across the entire routed optical topology. Hierarchical management considers infrastructure connections across geographically dispersed networks served by different operators. With standardized interfaces, such as Linux ONF T-API, monitoring support is provided for multivendor environments across different network domains or resource partitions.
The bottom line: MOFNs make optical networks ready for the AI-driven future
As AI continues to reshape digital infrastructure, optical networks grow in importance. Hyperscalers need the scale and flexibility of private builds, but with the local expertise, resources and regulatory compliance of CSP-operated networks to help reduce their overall risk to help grow their business. CSPs, meanwhile, need to improve their existing MOFN business model using automation to accelerate their delivery at hyperscaler speed while maintaining control and protection of their operational domains.
By adding a visibility layer, spectrum services and multivendor management – all powered and enhanced by network automation – CSPs can finally boost their MOFN business, increase its value against “dark fiber alternatives” or “managed services” to truly intelligent, high-assurance platforms that are ready to adhere to hyperscaler AI and cloud interconnect demands.
Read the full paper here.
