How flat is replacing fat in AWS data center networks

Summary

Traditional data center networks rely on hierarchical fat-tree topologies, which introduce overhead, upper-layer congestion, and vulnerability to router failures. Although flat random networks theoretically maximize route diversity and resilience, routing computation and cabling complexity historically made them impractical for physical data centers. AWS addressed these constraints by introducing Resilient Network Graphs, a quasi-random architecture powered by a passive optical device called ShuffleBox and a routing algorithm named Spraypoint. Spraypoint sprays packets across random neighbors toward designated waypoints, unlocking twice as many independent paths while operating within commodity router memory limits. Adopted as the default architecture across new AWS facilities, this design reduces router counts by 69%, lowers projected network power consumption by 40%, and increases throughput by up to 33%.

Context

Fat-tree data center architectures require extra router layers that add overhead, suffer from upper-layer congestion, and risk cutting off large network regions during single-router failures. Although flat random networks provide superior route diversity and graceful degradation, cabling them physically is infeasible, and computing multipath routing on random graphs requires 20 to 80 times more memory than commodity hardware provides.

Approach / What changed

AWS developed Resilient Network Graphs, a scalable flat quasi-random network architecture. The design uses ShuffleBox, a passive optical device that shuffles internal wires to enable structured physical cabling with quasi-random logical connections. For routing, the Spraypoint algorithm sprays packets to random neighbors and routes them through rings of waypoints to destinations, doubling independent paths while running on commodity router hardware.

Takeaways

• Resilient Network Graphs reduce the number of routers by 69%, achieve up to 33% higher throughput, and project a 40% reduction in network equipment electricity consumption compared to fat trees.
• The Spraypoint routing algorithm distributes traffic across random neighbors toward destination waypoints, delivering nearly twice as many independent paths as shortest-path routing on commodity router memory.
• The ShuffleBox passive optical device standardizes physical cabling to match fat-tree installation simplicity while producing a quasi-random logical topology.

Related reading

• Amazon · May 6, 2026

  Navigating uncertainty in Amazon's middle-mile network

  Amazon's middle-mile logistics network coordinates tens of millions of packages across hundreds of facilities while managing subtle demand variations and major operational disruptions. To solve this mixed-integer optimization challenge without combinatorial explosion, Amazon models routes at coarse time resolutions augmented by precomputed 15-minute timing bounds. The system uses Monte Carlo methods alongside a graph attention network featuring interconnected site graphs and origin-destination graphs to capture complex spatial correlations and flow interdependencies. By generating synthetic flow permutations, planners stress-test candidate network designs against hundreds of plausible scenarios to prioritize routing optionality. This risk-aware modeling approach preserves network flexibility, absorbs severe shocks, and yields potential cost savings while safeguarding delivery promises.

  Ruth Misener, Hana Ku, Georgios Paschos

• Meta · Aug 24, 2026