20/10/2025
Real-World Experience with 400G ZR+ and VXLAN: How We Achieved Seamless Connectivity Between Data Centers in Amsterdam
Why This Was Necessary
Amsterdam remains one of Europe’s largest colocation hubs. While there are plenty of data centers, the main limiting factor today is no longer rack space — it’s available power capacity.
Energy allocation rules are becoming increasingly strict, quotas are shrinking, and demand keeps growing.
As a result, companies are often forced to deploy their equipment wherever spare kilowatts are available — sometimes in adjacent halls, sometimes even in other data centers. For MIRhosting clients, it’s essential that all of this feels like a single, unified infrastructure, regardless of where the servers physically reside.
That’s why we built connectivity between sites using VXLAN overlays and 400G ZR+ optical modules.
Network Architecture: From Constraints to VXLAN
Previously, MIRhosting’s internal network operated at 100 G ports and relied on classical encapsulation schemes. As the number of services, provider connections, and total traffic grew, it became clear that a new architecture was needed — one that could scale horizontally, offer flexible segmentation, and distribute load seamlessly across sites.
To achieve this, we implemented VXLAN overlays — a technology that unites multiple sites into a single logical network and isolates segments without the limitations of traditional VLAN numbering or physical distribution.
As the core hardware, we chose Huawei CloudEngine CE9855-32DQ switches supporting 400 G ports and hardware-based VXLAN implementation.
This enabled us to move from a static 100 G topology to a flexible, scalable overlay/underlay network ready for future traffic growth and the onboarding of new sites.
Transport Layer: Transition to 400 G ZR+
Before the upgrade, our transport layer was built on DWDM solutions with 200/400 G transponders and multiplexers. The architecture was stable but overly complex for our use case — particularly when new links or capacity needed to be added quickly.
To simplify the infrastructure and increase efficiency, we adopted 400 G ZR+ optical modules installed directly into the core switch ports.
This approach removes the need for separate transponder chassis and simplifies operations while maintaining the required bandwidth.
Practical Aspects of 400 G ZR+
Working with 400 G ZR+ modules required careful attention to power and chassis design.
We used Huawei CloudEngine CE9865-4C switches — a 4U chassis with four interface cards.
For ZR+, only two of the four cards are used, and each of them supports only half of its ports. Thus, out of the nominal 32×400 G ports, eight are available for ZR+.
This limitation is due to the high power consumption and heat output of ZR+ optics, so part of the chassis capacity is reserved for stability and cooling.
It’s also important to note that ports configured for 400 G ZR+ are not compatible with 100 G modules — separate line cards or aggregation devices are used for those.
The resulting configuration is less dense, but it ensures stable operation and sufficient transmission distance for inter-DC connections.
What This Means for Clients
• Seamless connectivity between locations. Servers can reside in different racks, halls, or data centers in Amsterdam — and still operate within the same network.
• Scalable architecture. VXLAN enables flexible segmentation and readiness for further service and client growth.
• Minimal downtime. All upgrades were performed on the live network without interrupting client services.
• Predictable performance. 400 G ZR+ provides low latency and stability even under increased load.
Who Benefits from This Approach
• Providers and IT companies facing power and space constraints in Amsterdam.
• Organizations requiring L2/L3 transparency across multiple locations (for clustering, replication, or load balancing).
• Infrastructure teams seeking to reduce deployment time for new nodes and minimize dependence on inter-DC cross-connects.
In Conclusion
We didn’t just “add more racks.”
We made different locations work as one — despite Amsterdam’s growing power-capacity challenges.
For clients, this means the freedom to scale without compromising availability or performance.
If you’re looking for reliable and flexible colocation in Amsterdam — where the infrastructure is already built for inter-data-center connectivity and future growth — learn more about our Amsterdam colocation service https://mirhosting.com/colocation-in-amsterdam
