Building the Foundation for the Future: How Cyberlink Took Its Core Network to the Next Level

Cyberlink has completely modernized its core network: four new Cisco routers replace the previous hardware. The result is a tenfold increase in performance and an infrastructure built to last for years to come. This wasn't an ambitious stretch goal — it's a proven strategy. Cyberlink's engineering team doesn't plan reactively once the market starts applying pressure. The team plans years ahead. "Traffic doubles every two to three years," explains Senior Network Engineer Kim Graf. "We have to plan early and adapt our infrastructure to future demands before bottlenecks occur." This isn't just smart engineering — it's a promise to customers: premium infrastructure means stability through foresight.

The Challenge: Traffic That Never Stands Still

The decision to invest in new Cisco hardware was based on a simple but compelling realization: the old hardware had reached its limits. "Every five to seven years, we need to renew our infrastructure," explains Kim. "But not for the sake of the cycle — because demands keep growing." Many customers are seeing strong traffic growth. New customers are joining. Large customers are getting even larger. By onboarding new access networks, Cyberlink is reaching new industries and new geographic regions. The math simply stopped adding up.

Adding to the urgency, several hardware components were approaching end-of-support. "We would have had to renew the maintenance contracts," says Kim. "But if individual modules are going to lose support in just a few years, it makes no sense to keep running the complete hardware for another five years." The decision was made in early 2025: a complete replacement instead of incremental expansion.

Patrick Widmer, Product Manager at Cyberlink, points to another critical factor: "Highly scalable cloud workloads, inter-datacenter connections, and high-performance enterprise connectivity all demand ever-higher bandwidths. With the old hardware, our product development would have quickly hit a ceiling."

Tenfold More Performance — And Not Even at the Limit Yet

The new Cisco hardware delivers an impressive performance leap: instead of multiple 10-gigabit links between the core routers, the network can now carry multiple 100-gigabit links — a tenfold increase, available immediately. But that's only the beginning.

"What was decisive for us," says Kim, "is the modularity of the hardware. It's designed so we can upgrade to 400 gigabit without replacing the routers entirely. We simply add an additional module — like a Lego system. That's not just elegant; it will save us significant cost and effort five or six years from now. With the old hardware, we would have had to rip out the entire router and replace it with more expensive models. Now we can scale flexibly without rebuilding the infrastructure from the ground up."

Patrick adds a practical side effect: "The new hardware isn't just more powerful — it's also more compact. The old ASR 9006 took up considerably more rack space than the new solution. We'll use that freed-up space to expand new services. Our cloud teams need rack space for their infrastructure, and we in networking need room for additional systems. These resources are always scarce in a datacenter — with the new, more compact hardware, we gain valuable space for future investments."

The upgrade also opens up new possibilities on the sales side: services with 10-gigabit and higher bandwidths can now be marketed as standard offerings, without requiring detailed capacity checks beforehand. Previously, engineering had to verify for every customer with higher requirements whether sufficient bandwidth was still available on the core. If utilization was too high, requests had to be declined or services artificially capped. Going forward, the dynamic changes: the sales and product management teams can focus entirely on customer needs and business opportunities — the infrastructure now provides the capacity to reliably meet those demands. That's not just technically sound, it's commercially smart too: faster time-to-market, greater customer flexibility, stronger competitiveness.

Millimeter-Precision Work During Live Operations

A hardware swap in the datacenter looks simple from the outside — but it's a highly complex operation, as the details quickly reveal.

"Before the real engineering work even begins, a significant amount of effort is already required," says Kim. The physical installation is only the start.

The real challenge lies in the migration itself. It takes place during maintenance windows — at night and on weekends, when customer impact is minimal. Weeks of preparation precede it. Dozens of changes must be planned and coordinated. Central to this are the so-called test protocols: meticulous documentation that records, before the migration, which services are active and which parameters count as "normal."

"We write these protocols before the migration," explains Kim. "Then, after the migration, we work through them step by step — like a detailed checklist. At 4 a.m., after hours of work, you're grateful for that structure."

In concrete terms: the first migration took six hours. The following three took four to five hours each. Every service is tested and added to the new monitoring system. Beforehand, the team checks: does every customer have redundancy? Are there specific risks associated with this service?

The result speaks for itself: all four routers are now live, and the new infrastructure is ready for the decade ahead. "That fills me with pride," says Kim. "A project of this complexity, carried out during live operations, with a team that plans months in advance — that's no small achievement."

A Leaner Architecture, Higher Stability

Over the course of the project, the architecture was also reorganized. Historically, customer services and quality-of-service configurations had accumulated on the core network that didn't belong there. They belong on the distribution or access layer — not at the heart of the infrastructure.

"The core network should be like a main highway," explains Kim. "As lean as possible, as little configuration as possible, as stable as possible. Customer-specific configurations belong elsewhere." During the migration, these services were systematically moved outward. That makes the core network more stable, more performant, and easier to maintain.

The team also made a deliberate decision against automating the core configuration — which might sound counterintuitive, since automation is the dominant trend in the industry. But Kim has a clear rationale: "Once the core is set up, it needs minimal configuration changes. It doesn't make sense to build automation for something that might only be adjusted twice a year." Monitoring, telemetry, alerting, and capacity management, on the other hand, are fully automated — everything that runs on a regular basis.

What This Means for Customers

On the customer side, Patrick sums up the message concisely: "More capacity, higher performance, and long-term stability for business-critical applications."

In concrete terms: services with higher bandwidths become more affordable, because core capacity has grown massively. Customers with volatile demands can now benefit from services like "Burstable Internet Connect" — paying for one gigabit per second, but able to burst up to ten gigabits per second when needed. That was impossible with the old hardware; the core would have collapsed if many customers had burst simultaneously.

The systems remain just as reliable and high-performing as before. Through forward-looking planning of available resources, bottlenecks can be avoided early on. This creates more headroom before critical limits are reached — even if individual customers barely notice the difference day to day.

Ready for the Decade Ahead

With all four routers in operation, Kim and Patrick agree: Cyberlink is now optimally positioned for the years ahead. "We expect traffic to keep growing continuously," says Kim. "Whether that's driven by AI, streaming services, or cloud — that's hard to say. But our capacity is now so much larger that we can look at it with confidence."

Patrick adds: "At several locations, we've only just begun establishing a physical presence. Demand is high. That means we'll also grow horizontally — multiple cores across different locations. With the modular architecture of the new hardware, we can respond flexibly."

Above all, the project shows one thing: Cyberlink doesn't think in maintenance cycles — it thinks in lead time. Capacity is built before it's needed, not once the bottleneck has already arrived. For Cyberlink, that's more than a technical decision — it's part of who we are: premium infrastructure means stability through foresight. Something customers rarely notice consciously — but would feel immediately if it weren't there.