Service Provider

All in on multicloud connectivity

Lumen Technologies provides telecommunications services with extensive connectivity capabilities, including conduit, fiber, and connectivity transport services. A major network provider in the U.S., Lumen concentrates on enterprise business services, data center connectivity, data integration, managed hosting, cloud connectivity, and essential transport for communications. The rise of artificial intelligence (AI) and cloud-first strategies are reshaping operations and presenting evolving opportunities to create AI-driven processes and monetize differentiated services. For Lumen’s customer base—encompassing mid-market and large enterprises as well as public sector organizations at national, state, university, and research levels—bandwidth constraints must be addressed to support AI workloads.

The evolution of AI and its network connectivity requirements are challenging traditional internet architectures. Many existing networking architectures fail to account for the complexities of multicloud environments and AI applications. They’re not optimized for low-latency, high-load requirements and are limited in their ability to scale or efficiently allocate graphics processing unit (GPU) resources for AI. Additionally, many lack the flexibility to provide interconnectivity across cloud platforms and AI systems.

The centralized models and protocol inefficiencies of many existing architectures increase costs, power consumption, and latency, making them unsuitable for transporting AI workloads.

“We realized that internet architecture is not where it needs to be for an AI economy,” says Dave Ward, Chief Technology and Product Officer at Lumen. “We needed to fully upgrade our backbone to be able to meet the curve when bandwidth demands actually get above 1.6 terabits per second per link.”

Hear from Dave Ward, Lumen’s CTO, on evolving network architecture for the AI economy.

A backbone built for the AI landscape

Lumen’s leadership recognized another point of friction. “There’s an internet issue that we don’t talk about very much,” says Ward. “Carriers or service providers come together in something called carrier-neutral facilities.” These facilities were developed in the 1990s during the rise of the internet. In the U.S., there are only about a dozen such facilities spread around the country, and around the world the distribution is similar proportionally. All internet traffic flows through these facilities, whether it’s going to the cloud or cross-carrier. Moreover, carrier-neutral facilities are traditionally where hyperscalers connect to their customers.

Hyperscalers seeking sustainable power solutions as well as 100G+ wave services that are more consumable and application programming interface (API)-driven are influencing the emergence of new hubs. “When you’re in a carrier-neutral facility and connecting to multiple hyperscalers, you have to buy a direct connect to one cloud, an express route to another cloud, and ports—multiple ports—in all these locations,” says Ward. “The latency associated with moving from one cloud to another can’t possibly get any faster in today’s architecture than the links pulled to those carrier-neutral facilities.”

“With AI data centers and the AI economy really upon us, we realized we needed to be able to connect to those clouds and those data centers where they’re constructed—versus through these artificial constructs on the internet.” The cost of the architecture in a current multicloud environment is literally double the number of links and ports that are necessary. More efficient connections can create monetization opportunities, capitalizing on AI-driven network innovations to deliver differentiated services and assured experiences.

The company is developing custom solutions for this new, intelligent landscape through its Private Connectivity Fabric (PCF), which ensures seamless connectivity and API-driven consumption across all network layers. With Cisco Agile Services Networking, Lumen is transforming its network architecture and business service, using a more efficient network design for direct paths between multiple data centers and clouds.

Reducing the number of links and ports addresses the cost challenge head-on. “By changing the multicloud architecture away from the carrier-neutral facilities and away from these direct connects and express routes, you can save minimally 50%,” says Ward. “Then, if you take out cross connects, it’s over 60% savings—with a modern network architecture that makes data centers and hyperscalers first-class citizens.”

Expanding the architecture for tomorrow’s digital services

Leveraging this approach, Lumen has recreated its metro architecture, leaving behind the classical hub-and-spoke design around its metro areas. “We’ve based this architecture upon the Cisco NCS 1001, specifically built for metro routed optical networking,” says Ward. Lumen uses pluggable optics such as Cisco 400GE ZR/ZR+ coherent pluggable optics and dense Ethernet and IP routing equipment on the Cisco 8200 Series platform. With Routed Optical Networking from Cisco, this architecture reduces the cost of delivering a bit by 100%—and increases capacity in its fiber network by 1000 %.

“We’re now acquiring routers and switches and optical devices that are form, factor, and function fit for purpose for where we’re deploying them in the network. This catalyzes a critical change in the way we deliver enterprise business services,” says Ward. “In addition to upgrading our network to be able to handle 400G, 800G, and 1.6T connectivity, we have a direct interplay with the IP network for both public IP and our private IP or our enterprise business service offerings.”

Lumen is now equipped to not only deliver fiber to a data center but also offer an optical and routing solution to become the data center interconnect for new, multi-gigawatt data center parks. Moreover, Lumen has fundamentally expanded the control and automation of its network to include wavelengths, Ethernet, and IP. Control of endpoints and layers is driven by APIs and managed through routing protocols.

By using API-driven platforms for provisioning and telemetry, the architecture enables network changes and updates to be managed seamlessly across layers. “Being able to have planned capacity, planned redundancy, planned latency, and engineered bandwidth is really what’s fundamentally different in this architecture,” says Ward. “The fibers carry more optical channels than ever before—and we have the ability to automate and orchestrate across all three of those layers.”