{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:X2M5KM7H4FIU4PDWQEKRG2PPTX","short_pith_number":"pith:X2M5KM7H","schema_version":"1.0","canonical_sha256":"be99d533e7e1514e3c7681151369ef9dd69320cc5555f5011be8c90413bbe969","source":{"kind":"arxiv","id":"2410.01228","version":2},"attestation_state":"computed","paper":{"title":"ConServe: Fine-Grained GPU Harvesting for LLM Online and Offline Co-Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Haoran Ma, Harry Xu, Ion Stoica, Jiarong Xing, Joseph E. Gonzalez, Miryung Kim, Shan Yu, Shu Anzai, Shuo Yang, Yang Wang, Yang Zhou, Yifan Qiao, Yongji Wu","submitted_at":"2024-10-02T04:12:13Z","abstract_excerpt":"Large language model (LLM) serving demands low latency and high throughput, but high load variability makes it challenging to achieve high GPU utilization. In this paper, we identify a synergetic but overlooked opportunity to co-serve latency-critical online requests alongside latency-tolerant offline tasks such as model benchmarking. While promising, existing serving systems fail to co-serve them efficiently, as their coarse-grained resource management at the request or iteration level cannot harvest millisecond-level GPU idle cycles without introducing interference that violates online laten"},"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":"2410.01228","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2024-10-02T04:12:13Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"53eac75d18c5604b8c2b52a5ca4dc37cef56a503617cc560afa75b2b06c4ae74","abstract_canon_sha256":"741b0d30729a69c95d6f1b26e876ed13e63f02e003658aab873969b8f0ae450a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:04:28.478256Z","signature_b64":"jW+b+4jEwUfH2PwCjml/eNptvP0204KWndR7kT5eknyZd3W3O3vu8835up8TaNV3laBwGKCSqddJQvzadi9zDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be99d533e7e1514e3c7681151369ef9dd69320cc5555f5011be8c90413bbe969","last_reissued_at":"2026-07-05T12:04:28.477704Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:04:28.477704Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ConServe: Fine-Grained GPU Harvesting for LLM Online and Offline Co-Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Haoran Ma, Harry Xu, Ion Stoica, Jiarong Xing, Joseph E. Gonzalez, Miryung Kim, Shan Yu, Shu Anzai, Shuo Yang, Yang Wang, Yang Zhou, Yifan Qiao, Yongji Wu","submitted_at":"2024-10-02T04:12:13Z","abstract_excerpt":"Large language model (LLM) serving demands low latency and high throughput, but high load variability makes it challenging to achieve high GPU utilization. In this paper, we identify a synergetic but overlooked opportunity to co-serve latency-critical online requests alongside latency-tolerant offline tasks such as model benchmarking. While promising, existing serving systems fail to co-serve them efficiently, as their coarse-grained resource management at the request or iteration level cannot harvest millisecond-level GPU idle cycles without introducing interference that violates online laten"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.01228","kind":"arxiv","version":2},"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/2410.01228/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":"2410.01228","created_at":"2026-07-05T12:04:28.477765+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.01228v2","created_at":"2026-07-05T12:04:28.477765+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.01228","created_at":"2026-07-05T12:04:28.477765+00:00"},{"alias_kind":"pith_short_12","alias_value":"X2M5KM7H4FIU","created_at":"2026-07-05T12:04:28.477765+00:00"},{"alias_kind":"pith_short_16","alias_value":"X2M5KM7H4FIU4PDW","created_at":"2026-07-05T12:04:28.477765+00:00"},{"alias_kind":"pith_short_8","alias_value":"X2M5KM7H","created_at":"2026-07-05T12:04:28.477765+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05933","citing_title":"Beyond Greedy Chunking: SLO-Aware Sliding-Window Scheduling for LLM Inference","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28053","citing_title":"RW-TTT: Batched Serving for Request-Owned Test-Time Training State","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06534","citing_title":"ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06534","citing_title":"ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07609","citing_title":"Blink: CPU-Free LLM Inference by Delegating the Serving Stack to GPU and SmartNIC","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07874","citing_title":"Valve: Production Online-Offline Inference Colocation with Jointly-Bounded Preemption Latency and Rate","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06664","citing_title":"Foundry: Template-Based CUDA Graph Context Materialization for Fast LLM Serving Cold Start","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04745","citing_title":"The Energy Cost of Execution-Idle in GPU Clusters","ref_index":44,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX","json":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX.json","graph_json":"https://pith.science/api/pith-number/X2M5KM7H4FIU4PDWQEKRG2PPTX/graph.json","events_json":"https://pith.science/api/pith-number/X2M5KM7H4FIU4PDWQEKRG2PPTX/events.json","paper":"https://pith.science/paper/X2M5KM7H"},"agent_actions":{"view_html":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX","download_json":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX.json","view_paper":"https://pith.science/paper/X2M5KM7H","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.01228&json=true","fetch_graph":"https://pith.science/api/pith-number/X2M5KM7H4FIU4PDWQEKRG2PPTX/graph.json","fetch_events":"https://pith.science/api/pith-number/X2M5KM7H4FIU4PDWQEKRG2PPTX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX/action/storage_attestation","attest_author":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX/action/author_attestation","sign_citation":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX/action/citation_signature","submit_replication":"https://pith.science/pith/X2M5KM7H4FIU4PDWQEKRG2PPTX/action/replication_record"}},"created_at":"2026-07-05T12:04:28.477765+00:00","updated_at":"2026-07-05T12:04:28.477765+00:00"}