{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FX4GAA55ENWDTQQQ7F7RPJVSZ3","short_pith_number":"pith:FX4GAA55","schema_version":"1.0","canonical_sha256":"2df86003bd236c39c210f97f17a6b2cedf54f94e9fb34d61f6ceb531511ab764","source":{"kind":"arxiv","id":"2508.18572","version":1},"attestation_state":"computed","paper":{"title":"Strata: Hierarchical Context Caching for Long Context Language Model Serving","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Christos Kozyrakis, Mark Zhao, Michael Garland, Scott Mahlke, Vikram Sharma Mailthody, Yuwei An, Zhiqiang Xie, Ziyi Xu","submitted_at":"2025-08-26T00:09:03Z","abstract_excerpt":"Large Language Models (LLMs) with expanding context windows face significant performance hurdles. While caching key-value (KV) states is critical for avoiding redundant computation, the storage footprint of long-context caches quickly exceeds GPU memory capacity, forcing production systems to adopt hierarchical caching across memory hierarchies. However, transferring large cached contexts back to the GPU introduces severe performance bottlenecks: fragmented I/O from paged layouts prevents full bandwidth utilization, and existing schedulers fail to account for cache-loading delays, leaving syst"},"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":"2508.18572","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2025-08-26T00:09:03Z","cross_cats_sorted":[],"title_canon_sha256":"f8a52afef4404b3440c2205548f9ba8a90f7234f0beac2c4bcb7095c3582bb5c","abstract_canon_sha256":"b64c9c32b5362f2a8b6c7de9d464c185353864a9a4d479835132ef37fe58a838"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:59:26.946878Z","signature_b64":"SmBcpXut0Vxd4XQC2rkzILbuyHuUSNFFPSyAtgniEVJjQOC2Xwk6nq7+LOenYXLDG1zLQYMLS+M9xcF6TrM0Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2df86003bd236c39c210f97f17a6b2cedf54f94e9fb34d61f6ceb531511ab764","last_reissued_at":"2026-07-05T11:59:26.946414Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:59:26.946414Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Strata: Hierarchical Context Caching for Long Context Language Model Serving","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Christos Kozyrakis, Mark Zhao, Michael Garland, Scott Mahlke, Vikram Sharma Mailthody, Yuwei An, Zhiqiang Xie, Ziyi Xu","submitted_at":"2025-08-26T00:09:03Z","abstract_excerpt":"Large Language Models (LLMs) with expanding context windows face significant performance hurdles. While caching key-value (KV) states is critical for avoiding redundant computation, the storage footprint of long-context caches quickly exceeds GPU memory capacity, forcing production systems to adopt hierarchical caching across memory hierarchies. However, transferring large cached contexts back to the GPU introduces severe performance bottlenecks: fragmented I/O from paged layouts prevents full bandwidth utilization, and existing schedulers fail to account for cache-loading delays, leaving syst"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.18572","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/2508.18572/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":"2508.18572","created_at":"2026-07-05T11:59:26.946488+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.18572v1","created_at":"2026-07-05T11:59:26.946488+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.18572","created_at":"2026-07-05T11:59:26.946488+00:00"},{"alias_kind":"pith_short_12","alias_value":"FX4GAA55ENWD","created_at":"2026-07-05T11:59:26.946488+00:00"},{"alias_kind":"pith_short_16","alias_value":"FX4GAA55ENWDTQQQ","created_at":"2026-07-05T11:59:26.946488+00:00"},{"alias_kind":"pith_short_8","alias_value":"FX4GAA55","created_at":"2026-07-05T11:59:26.946488+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18071","citing_title":"KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19932","citing_title":"PEEK: Context Map as an Orientation Cache for Long-Context LLM Agents","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24391","citing_title":"FreqCache: Accelerating Embodied VLN Models with Adaptive Frequency-Guided Token Caching","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03375","citing_title":"Tutti: Making SSD-Backed KV Cache Practical for Long-Context LLM Serving","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06370","citing_title":"ForkKV: Scaling Multi-LoRA Agent Serving via Copy-on-Write Disaggregated KV Cache","ref_index":64,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3","json":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3.json","graph_json":"https://pith.science/api/pith-number/FX4GAA55ENWDTQQQ7F7RPJVSZ3/graph.json","events_json":"https://pith.science/api/pith-number/FX4GAA55ENWDTQQQ7F7RPJVSZ3/events.json","paper":"https://pith.science/paper/FX4GAA55"},"agent_actions":{"view_html":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3","download_json":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3.json","view_paper":"https://pith.science/paper/FX4GAA55","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.18572&json=true","fetch_graph":"https://pith.science/api/pith-number/FX4GAA55ENWDTQQQ7F7RPJVSZ3/graph.json","fetch_events":"https://pith.science/api/pith-number/FX4GAA55ENWDTQQQ7F7RPJVSZ3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3/action/storage_attestation","attest_author":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3/action/author_attestation","sign_citation":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3/action/citation_signature","submit_replication":"https://pith.science/pith/FX4GAA55ENWDTQQQ7F7RPJVSZ3/action/replication_record"}},"created_at":"2026-07-05T11:59:26.946488+00:00","updated_at":"2026-07-05T11:59:26.946488+00:00"}