{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RXRJRN674TJ4Z4WA776HG2H4EV","short_pith_number":"pith:RXRJRN67","schema_version":"1.0","canonical_sha256":"8de298b7dfe4d3ccf2c0fffc7368fc2564d33f9db4a62d2004a566d44b2c7905","source":{"kind":"arxiv","id":"2509.08309","version":1},"attestation_state":"computed","paper":{"title":"Hetis: Serving LLMs in Heterogeneous GPU Clusters with Fine-grained and Dynamic Parallelism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Chengzhong Xu, Huanle Xu, Jianxiong Liao, Zhi Zhou, Zizhao Mo","submitted_at":"2025-09-10T06:06:51Z","abstract_excerpt":"The significant resource demands in LLM serving prompts production clusters to fully utilize heterogeneous hardware by partitioning LLM models across a mix of high-end and low-end GPUs. However, existing parallelization approaches often struggle to scale efficiently in heterogeneous environments due to their coarse-grained and static parallelization strategies.\n  In this paper, we introduce Hetis, a new LLM system tailored for heterogeneous GPU clusters. Hetis addresses two critical challenges: (1) memory inefficiency caused by the mismatch between memory capacity and computational power in he"},"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":"2509.08309","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2025-09-10T06:06:51Z","cross_cats_sorted":[],"title_canon_sha256":"deca5f2cc298062ced1b837f6c3f75d7fd471d5c60759b80bb4d390e22453a03","abstract_canon_sha256":"81b08061b801198fff67062acaf61a4f9f38f1768d2e61a1e4c1d01f8f2dd8be"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:08:27.513805Z","signature_b64":"KMVHvy2DSfNOrhPNokp3SQcvdp7CX3EVOMXVHCnazBFURziMW9U6a5a1V0i5RE9Ug9vbjkXvS8PuVgKVTuxEDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8de298b7dfe4d3ccf2c0fffc7368fc2564d33f9db4a62d2004a566d44b2c7905","last_reissued_at":"2026-07-05T12:08:27.513343Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:08:27.513343Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hetis: Serving LLMs in Heterogeneous GPU Clusters with Fine-grained and Dynamic Parallelism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Chengzhong Xu, Huanle Xu, Jianxiong Liao, Zhi Zhou, Zizhao Mo","submitted_at":"2025-09-10T06:06:51Z","abstract_excerpt":"The significant resource demands in LLM serving prompts production clusters to fully utilize heterogeneous hardware by partitioning LLM models across a mix of high-end and low-end GPUs. However, existing parallelization approaches often struggle to scale efficiently in heterogeneous environments due to their coarse-grained and static parallelization strategies.\n  In this paper, we introduce Hetis, a new LLM system tailored for heterogeneous GPU clusters. Hetis addresses two critical challenges: (1) memory inefficiency caused by the mismatch between memory capacity and computational power in he"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.08309","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/2509.08309/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":"2509.08309","created_at":"2026-07-05T12:08:27.513417+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.08309v1","created_at":"2026-07-05T12:08:27.513417+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.08309","created_at":"2026-07-05T12:08:27.513417+00:00"},{"alias_kind":"pith_short_12","alias_value":"RXRJRN674TJ4","created_at":"2026-07-05T12:08:27.513417+00:00"},{"alias_kind":"pith_short_16","alias_value":"RXRJRN674TJ4Z4WA","created_at":"2026-07-05T12:08:27.513417+00:00"},{"alias_kind":"pith_short_8","alias_value":"RXRJRN67","created_at":"2026-07-05T12:08:27.513417+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.00946","citing_title":"Lodestar: An Online-Learning LLM Inference Router","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV","json":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV.json","graph_json":"https://pith.science/api/pith-number/RXRJRN674TJ4Z4WA776HG2H4EV/graph.json","events_json":"https://pith.science/api/pith-number/RXRJRN674TJ4Z4WA776HG2H4EV/events.json","paper":"https://pith.science/paper/RXRJRN67"},"agent_actions":{"view_html":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV","download_json":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV.json","view_paper":"https://pith.science/paper/RXRJRN67","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.08309&json=true","fetch_graph":"https://pith.science/api/pith-number/RXRJRN674TJ4Z4WA776HG2H4EV/graph.json","fetch_events":"https://pith.science/api/pith-number/RXRJRN674TJ4Z4WA776HG2H4EV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV/action/storage_attestation","attest_author":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV/action/author_attestation","sign_citation":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV/action/citation_signature","submit_replication":"https://pith.science/pith/RXRJRN674TJ4Z4WA776HG2H4EV/action/replication_record"}},"created_at":"2026-07-05T12:08:27.513417+00:00","updated_at":"2026-07-05T12:08:27.513417+00:00"}