{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:A4NOSSWTED4Z77Q3MAO6BS45WF","short_pith_number":"pith:A4NOSSWT","schema_version":"1.0","canonical_sha256":"071ae94ad320f99ffe1b601de0cb9db1630ff5ffee989759be71114548a260c7","source":{"kind":"arxiv","id":"2508.20274","version":1},"attestation_state":"computed","paper":{"title":"Predictable LLM Serving on GPU Clusters","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Erfan Darzi, Shreeanant Bharadwaj, Sree Bhargavi Balija","submitted_at":"2025-08-27T21:15:41Z","abstract_excerpt":"Latency-sensitive inference on shared A100 clusters often suffers noisy-neighbor interference on the PCIe fabric, inflating tail latency and SLO violations. We present a fabric-agnostic, VM-deployable host-level controller that combines dynamic Multi-Instance GPU (MIG) reconfiguration, PCIe-aware placement, and lightweight guardrails (MPS quotas, cgroup I/O). It samples per-tenant tails and system signals, uses topology hints to avoid PCIe hot spots, and gates actions with dwell/cool-down to avoid thrash. On a single host and a 2-node (16-GPU) cluster, SLO miss-rate is reduced by \\(\\approx\\)32"},"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":"2508.20274","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-08-27T21:15:41Z","cross_cats_sorted":[],"title_canon_sha256":"f20586fc94b94f9efc2778aa8b0113b78bab9ecc3799cfe848befaa8b1e16244","abstract_canon_sha256":"fac59c4df768227720a5ef8fb9c7cf91c118d4400e0805f91acf7338bfd1cad2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:00:33.416643Z","signature_b64":"igov9DsBHWT9pxNLrs4eFnps6QNo8/8y4oUsr7ZBXc6Ss0+6XXBC+M8HmnIw1BYjaIbR1HkVBaJETaA1PbssDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"071ae94ad320f99ffe1b601de0cb9db1630ff5ffee989759be71114548a260c7","last_reissued_at":"2026-07-05T12:00:33.416141Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:00:33.416141Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Predictable LLM Serving on GPU Clusters","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Erfan Darzi, Shreeanant Bharadwaj, Sree Bhargavi Balija","submitted_at":"2025-08-27T21:15:41Z","abstract_excerpt":"Latency-sensitive inference on shared A100 clusters often suffers noisy-neighbor interference on the PCIe fabric, inflating tail latency and SLO violations. We present a fabric-agnostic, VM-deployable host-level controller that combines dynamic Multi-Instance GPU (MIG) reconfiguration, PCIe-aware placement, and lightweight guardrails (MPS quotas, cgroup I/O). It samples per-tenant tails and system signals, uses topology hints to avoid PCIe hot spots, and gates actions with dwell/cool-down to avoid thrash. On a single host and a 2-node (16-GPU) cluster, SLO miss-rate is reduced by \\(\\approx\\)32"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.20274","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/2508.20274/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":"2508.20274","created_at":"2026-07-05T12:00:33.416201+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.20274v1","created_at":"2026-07-05T12:00:33.416201+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.20274","created_at":"2026-07-05T12:00:33.416201+00:00"},{"alias_kind":"pith_short_12","alias_value":"A4NOSSWTED4Z","created_at":"2026-07-05T12:00:33.416201+00:00"},{"alias_kind":"pith_short_16","alias_value":"A4NOSSWTED4Z77Q3","created_at":"2026-07-05T12:00:33.416201+00:00"},{"alias_kind":"pith_short_8","alias_value":"A4NOSSWT","created_at":"2026-07-05T12:00:33.416201+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF","json":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF.json","graph_json":"https://pith.science/api/pith-number/A4NOSSWTED4Z77Q3MAO6BS45WF/graph.json","events_json":"https://pith.science/api/pith-number/A4NOSSWTED4Z77Q3MAO6BS45WF/events.json","paper":"https://pith.science/paper/A4NOSSWT"},"agent_actions":{"view_html":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF","download_json":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF.json","view_paper":"https://pith.science/paper/A4NOSSWT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.20274&json=true","fetch_graph":"https://pith.science/api/pith-number/A4NOSSWTED4Z77Q3MAO6BS45WF/graph.json","fetch_events":"https://pith.science/api/pith-number/A4NOSSWTED4Z77Q3MAO6BS45WF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF/action/storage_attestation","attest_author":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF/action/author_attestation","sign_citation":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF/action/citation_signature","submit_replication":"https://pith.science/pith/A4NOSSWTED4Z77Q3MAO6BS45WF/action/replication_record"}},"created_at":"2026-07-05T12:00:33.416201+00:00","updated_at":"2026-07-05T12:00:33.416201+00:00"}