Data centre networking in transition: from classic architecture to AI-ready designs

AI-ready data centre networking does not simply describe a faster infrastructure, but an architectural approach in which automation, telemetry and intent-based operation are considered from the outset. In addition to AI data centres, these approaches and technological developments can also be used in general data centre infrastructures. This is precisely where Juniper Networks is positioning itself with its data centre networking portfolio.

Limits of classic data centre architectures

Many data centre networks in production today have grown historically. Typical features are

• Manual configuration of switches and routers
• Device-centred operation instead of a service- or application-oriented view
• Limited transparency in traffic flows and performance
• Static design assumptions that are based on predictable load profiles

These approaches quickly reach their limits in AI environments. Training clusters with GPUs or specialised accelerators generate massive, synchronous data streams. Even the smallest packet losses or microbursts can significantly extend the training duration or cause jobs to be cancelled completely. At the same time, the complexity of operations increases exponentially when networks have to be manually adapted or expanded.

New requirements due to AI and data-intensive workloads

AI workloads are fundamentally changing behaviour in the data centre:

• Significant east-west traffic between compute nodes
• High bandwidth requirements (100G, 400G and beyond)
• Ultra-low latency and deterministic behaviour
• Lossless or nearly lossless transport
• Dynamic scaling of clusters within the shortest possible time

A network must therefore not only be fast, but also predictable, adaptive and self-monitoring. This is precisely where an AI-ready architecture differs fundamentally from classic designs.

Architectural principles of modern AI-ready data centre fabrics (Juniper focus)

In contrast to traditional, often proprietary designs, Juniper Networks' modern approach is based on clearly defined, validated fabric architectures. These include in particular

• Spine-Leaf Fabrics (Layer 3 Clos Design) for horizontal scaling
• IP fabric with EVPN-VXLAN as an overlay for maximum flexibility
• Collapsed spine designs for smaller deployments
• Dedicated AI fabrics for GPU clusters with optimised east-west traffic

These designs enable deterministic performance and can scale linearly - a critical factor for AI workloads.

Intent-based networking with Juniper Apstra

A central element is the transition from device-based configuration to intent-based networking. Apstra assumes the position of central, multi-vendor SDN management in the data centre networking environment.

With Juniper Apstra, the desired state ("Intent") is defined, while the system automatically:

• Generates and rolls out configurations
• Validates dependencies between fabric components
• Performs continuous validation (closed loop)

This ensures that the network always corresponds exactly to the desired design - even after changes.

EVPN-VXLAN fabrics as a scalable basis

Modern data centre networks are based on spine-leaf architectures with EVPN-VXLAN as the underlay/overlay technology. Juniper extends this approach with deeply integrated automation and validation via Apstra as well as optimised control plane mechanisms. The complexity of the protocols used is thus abstracted by Apstra to such an extent that policies are rolled out in a service- and application-orientated, end-to-end manner.

advantages:

• Almost unlimited scalability
• Layer 2 and Layer 3 services across the entire fabric
• Clear separation of physical and logical topology
• Multi-tenant and multi-cluster capability (e.g. for applications, services, departments, business units, customers, ...)

Telemetry and AI-supported assurance with Juniper Mist AI

Instead of periodic SNMP polling, Juniper relies on real-time streaming telemetry.

In combination with Juniper Mist AI and Data Centre Assurance:

• End-to-end visibility across fabric, application and user behaviour
• Automatic root cause analysis (RCA)
• Proactive detection of performance degradation

Optimisation mechanisms for AI fabrics (RLB, DLB, ECMP enhancements)

Special fabric optimisation mechanisms are used for high-performance AI clusters:

• RLB (Random/Adaptive Load Balancing) for better distribution of flows
• DLB (Dynamic Load Balancing) to avoid hotspots
• Optimised ECMP hashing for even path utilisation
• Congestion avoidance mechanisms for low-loss transport

These technologies are crucial for minimising incast, microbursts and head-of-line blocking in GPU clusters.

Automation and lifecycle management

Another distinguishing feature compared to classic architectures is the fully automated lifecycle:

• Day-0: Design validation and simulation before commissioning
• Day-1: Zero-touch provisioning and automated fabric provisioning
• Day-2: Continuous validation, updates and expansions without interrupting operations

These functions are centrally controlled by Juniper Apstra and enable consistent operation over the entire lifecycle. A range of manufacturers can be integrated into Apstra. Brownfield, migration and multi-vendor scenarios can therefore also be realised.

Platforms and hardware: Juniper & HPE integration

Powerful switching platforms are an essential component of modern AI-ready architectures. Juniper relies in particular on the QFX series, including the Juniper QFX5250, for example.

These systems offer if required:

• 400G/800G interfaces for AI clusters
• Ultra-low latency switching
• High buffer capacities for bursty traffic

The collaboration between Hewlett Packard Enterprise (HPE) and Juniper results in integrated solutions in which compute, storage and networking are closely interlinked - especially for AI and HPC infrastructures. Traditional data centre network environments can also be implemented (e.g. 1/10/25/40/50/100GE).

Comparison: Today vs. AI-ready future

The following diagram provides an overview of the main differences between traditional and AI-ready data centre networking environments:

Conclusion: AI-ready data centre networking is not an evolutionary upgrade of existing networks, but a paradigm shift. The requirements of AI, ML and high-performance workloads cannot be met economically in the long term with traditional, manually operated architectures.

Juniper Networks' approach shows what modern data centre networks need to look like: automated, intent-based, transparent and intelligent. For companies that want to future-proof their infrastructure, this means not only a technological but also an organisational reorientation - away from pure network operation towards a software- and data-driven infrastructure model.

As a leading network solutions provider in Austria with in-depth expertise in network, data centre, application, automation and software development, CANCOM is the ideal partner to implement these projects with you.

This means that the network is no longer a limiting factor, but an accelerator for innovation and AI-driven business models.