{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KJKIMDXM2JINTYXPJLGDAC7K3P","short_pith_number":"pith:KJKIMDXM","schema_version":"1.0","canonical_sha256":"5254860eecd250d9e2ef4acc300beadbf61f7121e0558e60937589c2b8d4fdc2","source":{"kind":"arxiv","id":"2408.13510","version":2},"attestation_state":"computed","paper":{"title":"Intelligent Router for LLM Workloads: Improving Performance Through Workload-Aware Load Balancing","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.SY","eess.SY"],"primary_cat":"cs.DC","authors_text":"Anjaly Parayil, Ankur Mallick, Anoop Kulkarni, Chetan Bansal, Esha Choukse, \\'I\\~nigo Goiri, Jue Zhang, Kunal Jain, Rujia Wang, Saravan Rajmohan, Steve Kofsky, Victor R\\\"uhle, Xiaoting Qin","submitted_at":"2024-08-24T08:12:22Z","abstract_excerpt":"Large Language Model (LLM) workloads have distinct prefill and decode phases with different compute and memory requirements which should ideally be accounted for when scheduling input queries across different LLM instances in a cluster. However existing scheduling algorithms treat LLM workloads as monolithic jobs without considering the distinct characteristics of the two phases in each workload. This leads to sub-optimal scheduling and increased response latency. In this work, we start by characterizing factors affecting the response latency during LLM inference serving. We establish that bet"},"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":"2408.13510","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.DC","submitted_at":"2024-08-24T08:12:22Z","cross_cats_sorted":["cs.SY","eess.SY"],"title_canon_sha256":"8cb695fe2d65342e3d2d447c6a6ee5abd2fd903e5ae9d626b0e4bc28438e7760","abstract_canon_sha256":"87a466a189b42bb56073d888a15b11ef9501c1cf1500c4b5731b9fb9b31a161c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:58:11.193412Z","signature_b64":"O3CnMGX5GWB3R7j4yN+KCYwY62NvMWXob8wTEpbhk37DrB3uC0tP0Lbnl7a5quA42ZQwS2EwYdVAnRKoFH+vDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5254860eecd250d9e2ef4acc300beadbf61f7121e0558e60937589c2b8d4fdc2","last_reissued_at":"2026-07-05T09:58:11.192933Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:58:11.192933Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Intelligent Router for LLM Workloads: Improving Performance Through Workload-Aware Load Balancing","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.SY","eess.SY"],"primary_cat":"cs.DC","authors_text":"Anjaly Parayil, Ankur Mallick, Anoop Kulkarni, Chetan Bansal, Esha Choukse, \\'I\\~nigo Goiri, Jue Zhang, Kunal Jain, Rujia Wang, Saravan Rajmohan, Steve Kofsky, Victor R\\\"uhle, Xiaoting Qin","submitted_at":"2024-08-24T08:12:22Z","abstract_excerpt":"Large Language Model (LLM) workloads have distinct prefill and decode phases with different compute and memory requirements which should ideally be accounted for when scheduling input queries across different LLM instances in a cluster. However existing scheduling algorithms treat LLM workloads as monolithic jobs without considering the distinct characteristics of the two phases in each workload. This leads to sub-optimal scheduling and increased response latency. In this work, we start by characterizing factors affecting the response latency during LLM inference serving. We establish that bet"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.13510","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/2408.13510/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":"2408.13510","created_at":"2026-07-05T09:58:11.193002+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.13510v2","created_at":"2026-07-05T09:58:11.193002+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.13510","created_at":"2026-07-05T09:58:11.193002+00:00"},{"alias_kind":"pith_short_12","alias_value":"KJKIMDXM2JIN","created_at":"2026-07-05T09:58:11.193002+00:00"},{"alias_kind":"pith_short_16","alias_value":"KJKIMDXM2JINTYXP","created_at":"2026-07-05T09:58:11.193002+00:00"},{"alias_kind":"pith_short_8","alias_value":"KJKIMDXM","created_at":"2026-07-05T09:58:11.193002+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2504.09775","citing_title":"MIST: A Co-Design Framework for Heterogeneous, Multi-Stage LLM Inference","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":39,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P","json":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P.json","graph_json":"https://pith.science/api/pith-number/KJKIMDXM2JINTYXPJLGDAC7K3P/graph.json","events_json":"https://pith.science/api/pith-number/KJKIMDXM2JINTYXPJLGDAC7K3P/events.json","paper":"https://pith.science/paper/KJKIMDXM"},"agent_actions":{"view_html":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P","download_json":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P.json","view_paper":"https://pith.science/paper/KJKIMDXM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.13510&json=true","fetch_graph":"https://pith.science/api/pith-number/KJKIMDXM2JINTYXPJLGDAC7K3P/graph.json","fetch_events":"https://pith.science/api/pith-number/KJKIMDXM2JINTYXPJLGDAC7K3P/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P/action/storage_attestation","attest_author":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P/action/author_attestation","sign_citation":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P/action/citation_signature","submit_replication":"https://pith.science/pith/KJKIMDXM2JINTYXPJLGDAC7K3P/action/replication_record"}},"created_at":"2026-07-05T09:58:11.193002+00:00","updated_at":"2026-07-05T09:58:11.193002+00:00"}