{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:PRWHQLRDBXUMZPU2LM4IGO5V3W","short_pith_number":"pith:PRWHQLRD","schema_version":"1.0","canonical_sha256":"7c6c782e230de8ccbe9a5b38833bb5ddb7ba39b480147df81e0b9c08ee7b7f32","source":{"kind":"arxiv","id":"2109.11067","version":1},"attestation_state":"computed","paper":{"title":"Serving DNN Models with Multi-Instance GPUs: A Case of the Reconfigurable Machine Scheduling Problem","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Cheng Tan, Chuanxiong Guo, Jian Zhang, Sikai Qi, Yibo Zhu, Yu Cao, Zherui Liu, Zhichao Li","submitted_at":"2021-09-18T19:57:13Z","abstract_excerpt":"Multi-Instance GPU (MIG) is a new feature introduced by NVIDIA A100 GPUs that partitions one physical GPU into multiple GPU instances. With MIG, A100 can be the most cost-efficient GPU ever for serving Deep Neural Networks (DNNs). However, discovering the most efficient GPU partitions is challenging. The underlying problem is NP-hard; moreover, it is a new abstract problem, which we define as the Reconfigurable Machine Scheduling Problem (RMS). This paper studies serving DNNs with MIG, a new case of RMS. We further propose a solution, MIG-serving. MIG- serving is an algorithm pipeline that ble"},"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":"2109.11067","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2021-09-18T19:57:13Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ff7735271ab013c6e436dda1c275baeec4481a7fcc3ccb386a1a72f63736961a","abstract_canon_sha256":"450f256aab717f36bc72d5134559c74b9117b1835d5523992359cce47a71ed0c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:16:43.273588Z","signature_b64":"4VNNKOmJr8fCDNenKCg4vws156HYj2vKmcyHOWnun6jORg8NiJGIFEfn14JJ/uiWD6IkPeIOnuX4ifi9Mr2LDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7c6c782e230de8ccbe9a5b38833bb5ddb7ba39b480147df81e0b9c08ee7b7f32","last_reissued_at":"2026-07-05T03:16:43.273124Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:16:43.273124Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Serving DNN Models with Multi-Instance GPUs: A Case of the Reconfigurable Machine Scheduling Problem","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Cheng Tan, Chuanxiong Guo, Jian Zhang, Sikai Qi, Yibo Zhu, Yu Cao, Zherui Liu, Zhichao Li","submitted_at":"2021-09-18T19:57:13Z","abstract_excerpt":"Multi-Instance GPU (MIG) is a new feature introduced by NVIDIA A100 GPUs that partitions one physical GPU into multiple GPU instances. With MIG, A100 can be the most cost-efficient GPU ever for serving Deep Neural Networks (DNNs). However, discovering the most efficient GPU partitions is challenging. The underlying problem is NP-hard; moreover, it is a new abstract problem, which we define as the Reconfigurable Machine Scheduling Problem (RMS). This paper studies serving DNNs with MIG, a new case of RMS. We further propose a solution, MIG-serving. MIG- serving is an algorithm pipeline that ble"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2109.11067","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/2109.11067/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":"2109.11067","created_at":"2026-07-05T03:16:43.273178+00:00"},{"alias_kind":"arxiv_version","alias_value":"2109.11067v1","created_at":"2026-07-05T03:16:43.273178+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2109.11067","created_at":"2026-07-05T03:16:43.273178+00:00"},{"alias_kind":"pith_short_12","alias_value":"PRWHQLRDBXUM","created_at":"2026-07-05T03:16:43.273178+00:00"},{"alias_kind":"pith_short_16","alias_value":"PRWHQLRDBXUMZPU2","created_at":"2026-07-05T03:16:43.273178+00:00"},{"alias_kind":"pith_short_8","alias_value":"PRWHQLRD","created_at":"2026-07-05T03:16:43.273178+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25082","citing_title":"Energy Efficient Scheduling of AI/ML Workloads on Multi Instance GPUs with Dynamic Repartitioning","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30391","citing_title":"Energy-Aware Scheduling for Serverless LLM Serving on Shared GPUs","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29775","citing_title":"SMART-MIG: A Learning Framework for Scalable and Energy-Efficient GPU Scheduling","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01352","citing_title":"VUDA: Breaking CUDA-Vulkan Isolation for Spatial Sharing of Compute and Graphics on the Same GPU","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08451","citing_title":"Taming GPU Underutilization via Static Partitioning and Fine-grained CPU Offloading","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W","json":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W.json","graph_json":"https://pith.science/api/pith-number/PRWHQLRDBXUMZPU2LM4IGO5V3W/graph.json","events_json":"https://pith.science/api/pith-number/PRWHQLRDBXUMZPU2LM4IGO5V3W/events.json","paper":"https://pith.science/paper/PRWHQLRD"},"agent_actions":{"view_html":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W","download_json":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W.json","view_paper":"https://pith.science/paper/PRWHQLRD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2109.11067&json=true","fetch_graph":"https://pith.science/api/pith-number/PRWHQLRDBXUMZPU2LM4IGO5V3W/graph.json","fetch_events":"https://pith.science/api/pith-number/PRWHQLRDBXUMZPU2LM4IGO5V3W/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W/action/storage_attestation","attest_author":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W/action/author_attestation","sign_citation":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W/action/citation_signature","submit_replication":"https://pith.science/pith/PRWHQLRDBXUMZPU2LM4IGO5V3W/action/replication_record"}},"created_at":"2026-07-05T03:16:43.273178+00:00","updated_at":"2026-07-05T03:16:43.273178+00:00"}