{"paper":{"title":"RNG: Flat Datacenter Networks at Scale","license":"http://creativecommons.org/licenses/by/4.0/","headline":"RNG builds flat datacenter networks from quasi-random graphs that match fat-tree performance at up to 45 percent lower cost.","cross_cats":[],"primary_cat":"cs.NI","authors_text":"Chinchu Merine Joseph, C. Seshadhri, Elizabeth Tennent, Enrico Carlesso, Giacomo Bernardi, Luiza Popa, Pavan Manikonda, Randy Ram, Ratul Mahajan, Saurabh Kumar, Steven Robinson","submitted_at":"2026-04-16T17:37:04Z","abstract_excerpt":"We design and deploy in production the first flat datacenter networks. Our design, called RNG, is based on quasi-random graphs. While the cost and fault-tolerance benefits of such topologies have been long known, their practical realization has been hampered by a lack of scalable routing and cabling approaches. RNG has a new distributed routing protocol that exploits the properties of random graphs to find a large number of edge disjoint paths between pairs of endpoints. It uses a novel passive optical device that internally shuffles cables, which makes its cabling complexity similar to that o"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"RNG matches or exceeds the performance of fat trees for a range of traffic patterns, despite being up to 45% cheaper. RNG is now the default datacenter network for most workloads at Amazon.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The quasi-random graph topology combined with the new distributed routing protocol can consistently locate a large number of edge-disjoint paths at production scale, and the passive optical shuffling device introduces no new reliability or performance penalties in real deployments.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RNG is the first flat datacenter network deployed in production, based on quasi-random graphs with a scalable routing protocol and optical shuffling device, matching fat tree performance at up to 45% lower cost and now default at Amazon for most workloads.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"RNG builds flat datacenter networks from quasi-random graphs that match fat-tree performance at up to 45 percent lower cost.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"affd833e39fd2bcac0008fcdb11b3c56e16d5ce72872aac2b32be16a0e63ab2d"},"source":{"id":"2604.15261","kind":"arxiv","version":3},"verdict":{"id":"59126f7a-d59a-4489-8c36-13419a003b6e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T00:55:43.218946Z","strongest_claim":"RNG matches or exceeds the performance of fat trees for a range of traffic patterns, despite being up to 45% cheaper. RNG is now the default datacenter network for most workloads at Amazon.","one_line_summary":"RNG is the first flat datacenter network deployed in production, based on quasi-random graphs with a scalable routing protocol and optical shuffling device, matching fat tree performance at up to 45% lower cost and now default at Amazon for most workloads.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The quasi-random graph topology combined with the new distributed routing protocol can consistently locate a large number of edge-disjoint paths at production scale, and the passive optical shuffling device introduces no new reliability or performance penalties in real deployments.","pith_extraction_headline":"RNG builds flat datacenter networks from quasi-random graphs that match fat-tree performance at up to 45 percent lower cost."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.15261/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cf5a0b46892c86cd23461aff16bdbf8c823aea124e45d420c9dae9b43b1dffa0"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}