{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:4VTTZQ745LFO7ROJKXFQQLOX52","short_pith_number":"pith:4VTTZQ74","schema_version":"1.0","canonical_sha256":"e5673cc3fceacaefc5c955cb082dd7eea895c91160df220a8464914916a6ce46","source":{"kind":"arxiv","id":"2308.12908","version":1},"attestation_state":"computed","paper":{"title":"POLCA: Power Oversubscription in LLM Cloud Providers","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AR","cs.LG"],"primary_cat":"cs.DC","authors_text":"Brijesh Warrier, Chaojie Zhang, Esha Choukse, \\'I\\~nigo Goiri, Nithish Mahalingam, Pratyush Patel, Ricardo Bianchini","submitted_at":"2023-08-24T16:32:34Z","abstract_excerpt":"Recent innovation in large language models (LLMs), and their myriad use-cases have rapidly driven up the compute capacity demand for datacenter GPUs. Several cloud providers and other enterprises have made substantial plans of growth in their datacenters to support these new workloads. One of the key bottleneck resources in datacenters is power, and given the increasing model sizes of LLMs, they are becoming increasingly power intensive. In this paper, we show that there is a significant opportunity to oversubscribe power in LLM clusters. Power oversubscription improves the power efficiency of"},"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":"2308.12908","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.DC","submitted_at":"2023-08-24T16:32:34Z","cross_cats_sorted":["cs.AR","cs.LG"],"title_canon_sha256":"7d75fb49c705e1a9e644bf963782fa172496ecd1ef69b5319dc613c911561bc5","abstract_canon_sha256":"0190006db1f6c539c38e17b94d393280562e002eb9b9b0a156c88797967240ca"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:44:23.597833Z","signature_b64":"id527pRGeHS5sbO9H39dA09nwn+9x6HsINJ9rQ6QNyyntt8bUvPtzMbi6YiSEObXoeLVQmrK5+GOZoQDQCJCDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e5673cc3fceacaefc5c955cb082dd7eea895c91160df220a8464914916a6ce46","last_reissued_at":"2026-07-05T06:44:23.597227Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:44:23.597227Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"POLCA: Power Oversubscription in LLM Cloud Providers","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AR","cs.LG"],"primary_cat":"cs.DC","authors_text":"Brijesh Warrier, Chaojie Zhang, Esha Choukse, \\'I\\~nigo Goiri, Nithish Mahalingam, Pratyush Patel, Ricardo Bianchini","submitted_at":"2023-08-24T16:32:34Z","abstract_excerpt":"Recent innovation in large language models (LLMs), and their myriad use-cases have rapidly driven up the compute capacity demand for datacenter GPUs. Several cloud providers and other enterprises have made substantial plans of growth in their datacenters to support these new workloads. One of the key bottleneck resources in datacenters is power, and given the increasing model sizes of LLMs, they are becoming increasingly power intensive. In this paper, we show that there is a significant opportunity to oversubscribe power in LLM clusters. Power oversubscription improves the power efficiency of"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2308.12908","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/2308.12908/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":"2308.12908","created_at":"2026-07-05T06:44:23.597284+00:00"},{"alias_kind":"arxiv_version","alias_value":"2308.12908v1","created_at":"2026-07-05T06:44:23.597284+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2308.12908","created_at":"2026-07-05T06:44:23.597284+00:00"},{"alias_kind":"pith_short_12","alias_value":"4VTTZQ745LFO","created_at":"2026-07-05T06:44:23.597284+00:00"},{"alias_kind":"pith_short_16","alias_value":"4VTTZQ745LFO7ROJ","created_at":"2026-07-05T06:44:23.597284+00:00"},{"alias_kind":"pith_short_8","alias_value":"4VTTZQ74","created_at":"2026-07-05T06:44:23.597284+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19262","citing_title":"Detecting Hidden ML Training With Zero-Overhead Telemetry","ref_index":62,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21847","citing_title":"CompPow: A Case for Component-level GPU Power Management","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11999","citing_title":"The Illusion of Power Capping in LLM Decode: A Phase-Aware Energy Characterisation Across Attention Architectures","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01837","citing_title":"nvPAX: Constrained Optimization for Dynamic Power Allocation in Hierarchical and Multi-Tenant Systems","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52","json":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52.json","graph_json":"https://pith.science/api/pith-number/4VTTZQ745LFO7ROJKXFQQLOX52/graph.json","events_json":"https://pith.science/api/pith-number/4VTTZQ745LFO7ROJKXFQQLOX52/events.json","paper":"https://pith.science/paper/4VTTZQ74"},"agent_actions":{"view_html":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52","download_json":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52.json","view_paper":"https://pith.science/paper/4VTTZQ74","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2308.12908&json=true","fetch_graph":"https://pith.science/api/pith-number/4VTTZQ745LFO7ROJKXFQQLOX52/graph.json","fetch_events":"https://pith.science/api/pith-number/4VTTZQ745LFO7ROJKXFQQLOX52/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52/action/storage_attestation","attest_author":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52/action/author_attestation","sign_citation":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52/action/citation_signature","submit_replication":"https://pith.science/pith/4VTTZQ745LFO7ROJKXFQQLOX52/action/replication_record"}},"created_at":"2026-07-05T06:44:23.597284+00:00","updated_at":"2026-07-05T06:44:23.597284+00:00"}