{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UKNAV3CEW7FQHCMW3Z44T2CB3S","short_pith_number":"pith:UKNAV3CE","schema_version":"1.0","canonical_sha256":"a29a0aec44b7cb038996de79c9e841dc862422421882c6c7a2307a4446d1264d","source":{"kind":"arxiv","id":"2607.28633","version":1},"attestation_state":"computed","paper":{"title":"Topology-Aware Data Movement for Disaggregated GPU Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.PF"],"primary_cat":"cs.LG","authors_text":"Sanjeev Rao Ganjihal","submitted_at":"2026-04-19T22:46:57Z","abstract_excerpt":"Disaggregated LLM inference creates a datacenter networking problem that no existing system solves correctly. When prefill and decode run on separate GPU pools, the KV cache must be transferred between them. For a 70B model this is 2.6 GB per request, exceeding 100 GB/s aggregate at production scale. Yet DistServe, Splitwise, and Mooncake all use uniform RDMA, ignoring that bandwidth between two GPUs varies by 72x depending on their physical relationship: 900 GB/s via NVLink within a domain, 50 GB/s via InfiniBand across nodes, 12.5 GB/s via TCP across data centers. We design a topology-aware "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2607.28633","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-19T22:46:57Z","cross_cats_sorted":["cs.AI","cs.PF"],"title_canon_sha256":"09977caa6cccc417d18fa6438ed0546abe65b52c97711c2ea855943ad7a70a1a","abstract_canon_sha256":"a009543d0379ec75ebf93d9b90642341af37163aad3526ba7c3c2d8c4bf0d41b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-03T00:11:51.630648Z","signature_b64":"EFdoTF/PWr6Kf5P39LRnqmXnQGCEgTcQ0+L0KAKZnAdoN4Hs/ea4wm/yrqwrk4KxIph9f7XnRhjMwR1PnDE4Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a29a0aec44b7cb038996de79c9e841dc862422421882c6c7a2307a4446d1264d","last_reissued_at":"2026-08-03T00:11:51.628517Z","signature_status":"signed_v1","first_computed_at":"2026-08-03T00:11:51.628517Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Topology-Aware Data Movement for Disaggregated GPU Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.PF"],"primary_cat":"cs.LG","authors_text":"Sanjeev Rao Ganjihal","submitted_at":"2026-04-19T22:46:57Z","abstract_excerpt":"Disaggregated LLM inference creates a datacenter networking problem that no existing system solves correctly. When prefill and decode run on separate GPU pools, the KV cache must be transferred between them. For a 70B model this is 2.6 GB per request, exceeding 100 GB/s aggregate at production scale. Yet DistServe, Splitwise, and Mooncake all use uniform RDMA, ignoring that bandwidth between two GPUs varies by 72x depending on their physical relationship: 900 GB/s via NVLink within a domain, 50 GB/s via InfiniBand across nodes, 12.5 GB/s via TCP across data centers. We design a topology-aware "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.28633","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2607.28633/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":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2607.28633","created_at":"2026-08-03T00:11:51.630026+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.28633v1","created_at":"2026-08-03T00:11:51.630026+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.28633","created_at":"2026-08-03T00:11:51.630026+00:00"},{"alias_kind":"pith_short_12","alias_value":"UKNAV3CEW7FQ","created_at":"2026-08-03T00:11:51.630026+00:00"},{"alias_kind":"pith_short_16","alias_value":"UKNAV3CEW7FQHCMW","created_at":"2026-08-03T00:11:51.630026+00:00"},{"alias_kind":"pith_short_8","alias_value":"UKNAV3CE","created_at":"2026-08-03T00:11:51.630026+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S","json":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S.json","graph_json":"https://pith.science/api/pith-number/UKNAV3CEW7FQHCMW3Z44T2CB3S/graph.json","events_json":"https://pith.science/api/pith-number/UKNAV3CEW7FQHCMW3Z44T2CB3S/events.json","paper":"https://pith.science/paper/UKNAV3CE"},"agent_actions":{"view_html":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S","download_json":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S.json","view_paper":"https://pith.science/paper/UKNAV3CE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.28633&json=true","fetch_graph":"https://pith.science/api/pith-number/UKNAV3CEW7FQHCMW3Z44T2CB3S/graph.json","fetch_events":"https://pith.science/api/pith-number/UKNAV3CEW7FQHCMW3Z44T2CB3S/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S/action/storage_attestation","attest_author":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S/action/author_attestation","sign_citation":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S/action/citation_signature","submit_replication":"https://pith.science/pith/UKNAV3CEW7FQHCMW3Z44T2CB3S/action/replication_record"}},"created_at":"2026-08-03T00:11:51.630026+00:00","updated_at":"2026-08-03T00:11:51.630026+00:00"}