Network Resilience as a Key Pillar of Modern Infrastructures

Resilient networks are a central prerequisite for the operation of modern data centers and telecommunications infrastructures. Increasing demands from artificial intelligence (AI) applications, growing data volumes, and new security threats are placing greater emphasis on network architecture and network operations.

In an interview with eco, Matthias Hahn, Account Management/Network Infrastructure at Nokia, discusses current challenges for data centers and telecommunications networks, technological developments beyond pure bandwidth, and the importance of quantum-secure encryption for long-term network planning. 

dotmagazine: Resilience is often defined in technical terms. When you look at data centers and telecommunications networks, where do you see the biggest systemic weaknesses today, not so much in the hardware, but in the design or operation of networks?

Matthias: If hardware and software come from reliable sources, the weaknesses are indeed more likely to lie in the design and operation of networks. Redundant entry points to and within the data center, as well as disjoint routing in the wide area network outside the data center, are essential. Encryption of data in transit between data centers and toward customers, as well as protection against DDoS attacks, are standard today, or at least should be.

AI workloads are currently changing network requirements significantly. What new pressures do you see in particular for data center backbones? And where do classic architectural principles reach their limits?

Matthias: At present, AI workloads are processed in a very confined space, typically within GPU clusters, for example within a single rack. When workloads are distributed across multiple racks in a scale-out scenario, for instance within a data center section, this already has a major impact on the design of the AI backend. 

The effects will be even more pronounced when scale-across approaches distribute workloads across multiple sections or even different data center sites. The industry is currently developing the appropriate networking technologies for this. For example, many major players in the networking sector are jointly advancing the new Ultra Ethernet standard. In addition, rail-based approaches are increasingly used in AI backends, in some cases offering significant advantages over leaf–spine architectures. A key challenge is often integrating these approaches into a homogeneous operational model for the entire data center.

WDM and IP networks are considered mature technologies. In which areas are you currently seeing the greatest leaps in innovation? And which of these are truly relevant for network resilience, beyond pure bandwidth increases?

Matthias: Bandwidth increases are, of course, essential. Time and again, it is striking how quickly bandwidths that initially seemed oversized are actually required. In view of the expected spread of commercially viable quantum computers, practical encryption solutions for the “post-quantum era” are also becoming increasingly important.

Network operations are becoming ever more complex: frequent SW updates, integration of new systems into existing environments, and rapid data center expansion at unprecedented speed. At the same time, this growing complexity coincides with a shortage of trained network engineers. For this reason, the greatest need for innovation lies in network automation. Automation has been discussed for years without any substantial progress. We now appear to have reached a point where it is recognized that, without genuine automation, it will no longer be possible to operate complex network infrastructures securely and without errors. Planning with digital twins, continuous actual–target comparisons, and comprehensive rollback scenarios are likely to become standard within a few years, enabling networks to be operated economically and securely. At Nokia and among our partners, this growing complexity is also reflected in the increasing number of requests for design and architecture workshops.

Security and resilience are often considered separately. How is the debate around quantum-secure encryption changing perspectives on long-term network planning, particularly for operators of critical infrastructures?
Matthias: 

Operators of critical infrastructures are already under constant attack and must effectively defend themselves against DDoS attacks and social engineering. Quantum-secure encryption is an important building block in network planning, helping to prevent direct interception of data during transmission, for example via fiber-optic couplers. The greater an organization’s need for protection, the more network layers it will encrypt. From a technical perspective, this is already feasible today at both optical and IP layers.

The panel discussion focuses on solutions for the data center and telecommunications sectors. If you had to name three key questions decision-makers should be asking themselves today to make their networks future-proof and resilient, what would they be?
Matthias: In my view, the following three questions are particularly central:

1. Is network operation sufficiently automated, flexible, and scalable to meet future requirements? Are human input errors, for example resulting from non-automated work directly on the command-line interface of network elements, effectively excluded?
2. Do you really know which path your data takes, and is this path transparent and traceable?
3. Is the network design prepared for rapidly changing requirements at new locations and for shifting workloads?

Those who can answer these questions positively are laying the foundation for a network that not only safeguards current operations, but also supports long-term business requirements.

With more than 25 years of experience in the telecommunications industry, Matthias Hahn has a deep understanding of the needs and challenges of network operators.

He began his career in radio network engineering, where he held several leadership positions, before moving into more customer‑focused roles in service development and later in sales.

In his current sales role, he concentrates on all networking aspects for data center and cloud providers, with a strong focus on enabling the AI era.