AWS Replaces Fat-Tree Data Center Networks with Random Graph Theory, Cutting Routers by 69%

2026-06-04 · Source: InfoQ · Field: Technology & Digital — Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

AWS has adopted Resilient Network Graphs (RNG), a flat network architecture based on quasi-random graph theory, as the default for most new non-GPU data center builds globally as of April 2026. This marks the first large-scale production deployment of expander-based network fabrics, replacing traditional fat-tree topologies. The RNG approach, detailed in an arXiv paper, achieves 69% fewer networking devices, up to 33% higher throughput, and a projected 40% reduction in network equipment power consumption. It eliminates hierarchical spine and leaf layers, connecting top-of-rack switches directly via passive optical ShuffleBoxes for quasi-random physical links and using a custom Spraypoint protocol for distributed routing. This design ensures proportional network degradation rather than catastrophic failure. Validation included 530 processor-years of simulation, with initial deployments in Ireland, Germany, and Spain starting in late 2024.

Key takeaway

For Network Architects evaluating next-generation data center designs, AWS's adoption of Resilient Network Graphs (RNG) demonstrates a viable path to significantly reduce hardware and power consumption. You should investigate expander-based network fabrics as an alternative to traditional fat-tree topologies, especially for general-purpose compute workloads. Consider how a flat, quasi-random network could improve your infrastructure's resilience and throughput, potentially cutting device count by over 60%.

Key insights

Random graph theory applied to data center networks significantly reduces hardware while improving resilience and throughput.

Principles

• Random connectivity optimizes network efficiency.
• Passive optical devices simplify complex physical topologies.
• Distributed routing enhances fault tolerance.

Method

AWS implemented RNG by replacing spine/leaf layers with a mesh of ToR switches connected via ShuffleBoxes for quasi-random links and using Spraypoint, a custom distributed routing protocol, to guide packets.

In practice

• Evaluate expander-based network fabrics for general compute.
• Consider passive optical devices for physical layer randomization.
• Design routing protocols for flat, distributed topologies.

Topics

• Resilient Network Graphs
• Data Center Networks
• Random Graph Theory
• Network Topology
• ShuffleBox
• Spraypoint

Best for: CTO, VP of Engineering/Data, MLOps Engineer, AI Architect, DevOps Engineer, Research Scientist

Related on AIssential

• How flat is replacing fat in AWS data center networks — AWS's RNG flat network design, using quasi-random topologies and ShuffleBoxes, significantly improves data center efficiency and resilience.
• Mapping the Network: A Beginner’s Guide to Graph Theory and Python — Graph theory models relationships between objects using nodes and edges to optimize network understanding.
• Graph Rewiring in GNNs to Mitigate Over-Squashing and Over-Smoothing: A Survey — Graph rewiring modifies GNN topology to combat over-squashing and over-smoothing, improving information flow.
• Achieving the Kesten-Stigum bound in the non-uniform hypergraph stochastic block model — Combining subcritical hypergraph layers can enable weak recovery if their aggregate signal-to-noise ratio surpasses a Kesten–Stigum-type bound.
• Optimization of Edge Directions and Weights for Mixed Guidance Graphs in Lifelong Multi-Agent Path Finding — Optimizing both edge directions and weights in guidance graphs significantly improves LMAPF throughput and collision avoidance.
• AdaGraph: A Graph-Native Clustering Algorithm That Overcomes the Curse of Dimensionality and Enables Scientific Discovery — Structure-Centric Machine Learning (SC-ML) and AdaGraph overcome the curse of dimensionality by using graph topology instead of Euclidean distances.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by InfoQ.