{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ALPU2AHEDAD5SHTVHOVAZMII7Q","short_pith_number":"pith:ALPU2AHE","schema_version":"1.0","canonical_sha256":"02df4d00e41807d91e753baa0cb108fc1eaa3ab45cabefbe620ebe23e93ecb8f","source":{"kind":"arxiv","id":"2509.04827","version":3},"attestation_state":"computed","paper":{"title":"VoltanaLLM: Energy-Efficient and SLO-Aware Disaggregated LLM Serving via Adaptive Frequency Control and State-Space Routing","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Aryan Taneja, Jiahuan Yu, Junfeng Lin, Minjia Zhang","submitted_at":"2025-09-05T05:58:16Z","abstract_excerpt":"The energy cost of Large Language Model (LLM) inference is rapidly becoming a barrier to sustainable and scalable deployment. Although modern serving architectures expose distinct prefill and decode behaviors, existing systems fail to exploit these phase differences for energy-efficient serving under strict latency SLOs. This paper introduces VoltanaLLM, the first system that explicitly targets and reduces the energy bloat in modern prefill-decode (P/D) disaggregated LLM serving. Guided by a control-theory perspective, VoltanaLLM separates two levers: per-instance operating-point selection (GP"},"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":"2509.04827","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.DC","submitted_at":"2025-09-05T05:58:16Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2d3092824e4104fbc35d6aa80a1a62268ab52a8e502e81deefd1692e601f77c9","abstract_canon_sha256":"c0556c0b1a073d5085ad7ab016b69d81559d366ae039db5a5e2b8b24805b2195"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-24T00:14:20.870403Z","signature_b64":"fBiwAKhzJrMGwDTpHlmrwEU4vJFfDvYrZo9fmUXmQPTZFD6Qe5K6UWQabZ0NFeKJC6r6cgDQASQh3ufnaPiPCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"02df4d00e41807d91e753baa0cb108fc1eaa3ab45cabefbe620ebe23e93ecb8f","last_reissued_at":"2026-06-24T00:14:20.869901Z","signature_status":"signed_v1","first_computed_at":"2026-06-24T00:14:20.869901Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"VoltanaLLM: Energy-Efficient and SLO-Aware Disaggregated LLM Serving via Adaptive Frequency Control and State-Space Routing","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Aryan Taneja, Jiahuan Yu, Junfeng Lin, Minjia Zhang","submitted_at":"2025-09-05T05:58:16Z","abstract_excerpt":"The energy cost of Large Language Model (LLM) inference is rapidly becoming a barrier to sustainable and scalable deployment. Although modern serving architectures expose distinct prefill and decode behaviors, existing systems fail to exploit these phase differences for energy-efficient serving under strict latency SLOs. This paper introduces VoltanaLLM, the first system that explicitly targets and reduces the energy bloat in modern prefill-decode (P/D) disaggregated LLM serving. Guided by a control-theory perspective, VoltanaLLM separates two levers: per-instance operating-point selection (GP"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.04827","kind":"arxiv","version":3},"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/2509.04827/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":"2509.04827","created_at":"2026-06-24T00:14:20.869959+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.04827v3","created_at":"2026-06-24T00:14:20.869959+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.04827","created_at":"2026-06-24T00:14:20.869959+00:00"},{"alias_kind":"pith_short_12","alias_value":"ALPU2AHEDAD5","created_at":"2026-06-24T00:14:20.869959+00:00"},{"alias_kind":"pith_short_16","alias_value":"ALPU2AHEDAD5SHTV","created_at":"2026-06-24T00:14:20.869959+00:00"},{"alias_kind":"pith_short_8","alias_value":"ALPU2AHE","created_at":"2026-06-24T00:14:20.869959+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2606.30391","citing_title":"Energy-Aware Scheduling for Serverless LLM Serving on Shared GPUs","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2605.23348","citing_title":"XWind: A Cross-site Router for Large Language Model Inference Serving at Renewable Energy Farms","ref_index":83,"is_internal_anchor":true},{"citing_arxiv_id":"2601.20309","citing_title":"SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips","ref_index":23,"is_internal_anchor":true},{"citing_arxiv_id":"2604.16682","citing_title":"KAIROS: Stateful, Context-Aware Power-Efficient Agentic Inference Serving","ref_index":74,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q","json":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q.json","graph_json":"https://pith.science/api/pith-number/ALPU2AHEDAD5SHTVHOVAZMII7Q/graph.json","events_json":"https://pith.science/api/pith-number/ALPU2AHEDAD5SHTVHOVAZMII7Q/events.json","paper":"https://pith.science/paper/ALPU2AHE"},"agent_actions":{"view_html":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q","download_json":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q.json","view_paper":"https://pith.science/paper/ALPU2AHE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.04827&json=true","fetch_graph":"https://pith.science/api/pith-number/ALPU2AHEDAD5SHTVHOVAZMII7Q/graph.json","fetch_events":"https://pith.science/api/pith-number/ALPU2AHEDAD5SHTVHOVAZMII7Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q/action/storage_attestation","attest_author":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q/action/author_attestation","sign_citation":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q/action/citation_signature","submit_replication":"https://pith.science/pith/ALPU2AHEDAD5SHTVHOVAZMII7Q/action/replication_record"}},"created_at":"2026-06-24T00:14:20.869959+00:00","updated_at":"2026-06-24T00:14:20.869959+00:00"}