{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UPSKCNFDPIEPWTBSKLPJNT5O4A","short_pith_number":"pith:UPSKCNFD","schema_version":"1.0","canonical_sha256":"a3e4a134a37a08fb4c3252de96cfaee015782c6dd8e4c64a70f170f11a3e38da","source":{"kind":"arxiv","id":"2407.00079","version":4},"attestation_state":"computed","paper":{"title":"Mooncake: A KVCache-centric Disaggregated Architecture for LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.DC","authors_text":"Mingxing Zhang, Ruoyu Qin, Weimin Zheng, Weiran He, Xinran Xu, Yongwei Wu, Zheming Li","submitted_at":"2024-06-24T02:05:32Z","abstract_excerpt":"Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI. It features a KVCache-centric disaggregated architecture that separates the prefill and decoding clusters. It also leverages the underutilized CPU, DRAM, and SSD resources of the GPU cluster to implement a disaggregated cache of KVCache. The core of Mooncake is its KVCache-centric scheduler, which balances maximizing overall effective throughput while meeting latency-related Service Level Objectives (SLOs). Unlike traditional studies that assume all requests will be processed, Mooncake faces challenges du"},"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":"2407.00079","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2024-06-24T02:05:32Z","cross_cats_sorted":["cs.AI","cs.AR"],"title_canon_sha256":"30655276816209d843fae689c644cc400fb4e386467c6bb672381316945c0186","abstract_canon_sha256":"870dd1cafaa8bcf598ccb189c657338394d55bc7429bd11b5a3dbc5cfc62751a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:04:12.860629Z","signature_b64":"ZorMFmQRbTfBvjV9iFfa2LGnD4HmOTryy7aAZfJKZxDcCWqcWNiIRgALbJut5JdDHzY5QTgAk/EIMtVcXPzOCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3e4a134a37a08fb4c3252de96cfaee015782c6dd8e4c64a70f170f11a3e38da","last_reissued_at":"2026-07-05T12:04:12.860044Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:04:12.860044Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mooncake: A KVCache-centric Disaggregated Architecture for LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.DC","authors_text":"Mingxing Zhang, Ruoyu Qin, Weimin Zheng, Weiran He, Xinran Xu, Yongwei Wu, Zheming Li","submitted_at":"2024-06-24T02:05:32Z","abstract_excerpt":"Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI. It features a KVCache-centric disaggregated architecture that separates the prefill and decoding clusters. It also leverages the underutilized CPU, DRAM, and SSD resources of the GPU cluster to implement a disaggregated cache of KVCache. The core of Mooncake is its KVCache-centric scheduler, which balances maximizing overall effective throughput while meeting latency-related Service Level Objectives (SLOs). Unlike traditional studies that assume all requests will be processed, Mooncake faces challenges du"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.00079","kind":"arxiv","version":4},"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/2407.00079/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":"2407.00079","created_at":"2026-07-05T12:04:12.860110+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.00079v4","created_at":"2026-07-05T12:04:12.860110+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.00079","created_at":"2026-07-05T12:04:12.860110+00:00"},{"alias_kind":"pith_short_12","alias_value":"UPSKCNFDPIEP","created_at":"2026-07-05T12:04:12.860110+00:00"},{"alias_kind":"pith_short_16","alias_value":"UPSKCNFDPIEPWTBS","created_at":"2026-07-05T12:04:12.860110+00:00"},{"alias_kind":"pith_short_8","alias_value":"UPSKCNFD","created_at":"2026-07-05T12:04:12.860110+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":36,"internal_anchor_count":6,"sample":[{"citing_arxiv_id":"2607.05876","citing_title":"Think Before You Grid-Search: Floor-First Triage for LLM Serving","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2607.06519","citing_title":"FreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM Inference","ref_index":103,"is_internal_anchor":true},{"citing_arxiv_id":"2607.08032","citing_title":"What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents","ref_index":96,"is_internal_anchor":true},{"citing_arxiv_id":"2607.08057","citing_title":"Towards Efficient Large Language Model Serving: A Survey on System-Aware KV Cache Optimization","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2607.05876","citing_title":"Think Before You Grid-Search: Floor-First Triage for LLM Serving","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2607.06523","citing_title":"DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression","ref_index":117,"is_internal_anchor":true},{"citing_arxiv_id":"2606.20295","citing_title":"Token-Operations-Oriented Inference Optimization Techniques for Large Models","ref_index":176,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00466","citing_title":"ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11690","citing_title":"Beyond Per-Token Pricing: A Concurrency-Aware Methodology for LLM Infrastructure Cost Estimation","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01579","citing_title":"OmniPilot: An Uncertainty-Aware LLM Inference Advisor for Heterogeneous GPU Clusters","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08635","citing_title":"SpectrumKV: Per-Token Mixed-Precision KV Cache Transfer for Prefill-Decode Disaggregated LLM Serving","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07684","citing_title":"Semantic Cache Distillation: Efficient State Transfer via Reuse and Selective Patching","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06453","citing_title":"Vortex: Efficient and Programmable Sparse Attention Serving for AI Agents","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00466","citing_title":"ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00866","citing_title":"Idleness is Relative: Exploiting Tool-Call Idle Windows for Offloading in Agentic Systems with MORI","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00288","citing_title":"Model-Native Computing Architecture: Envisioning Future System Architecture Through the Lens of Computer Architecture","ref_index":128,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31093","citing_title":"Omni-Flow: A Unified Workflow Orchestration and Distributed KV Cache Sharing Framework for Multimodal Inference","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29708","citing_title":"Demystifying the Design Space and Best Practices for Heterogeneous LLM Inference and Serving","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20577","citing_title":"Human-Less LLM Serving: Quantifying the Human Tax on Throughput","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29708","citing_title":"Demystifying the Design Space and Best Practices for Heterogeneous LLM Inference and Serving","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2412.03594","citing_title":"BatchLLM: Optimizing Large Batched LLM Inference with Global Prefix Sharing and Throughput-oriented Token Batching","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2507.18454","citing_title":"Sandwich: Joint Configuration Search and Hot-Switching for Efficient CPU LLM Serving","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2510.18586","citing_title":"TokenCake: A KV-Cache-centric Serving Framework for LLM-based Multi-Agent Applications","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2505.23970","citing_title":"Cache Your Prompt When It's Green: Carbon-Aware Caching for Large Language Model Serving","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2502.10248","citing_title":"Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A","json":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A.json","graph_json":"https://pith.science/api/pith-number/UPSKCNFDPIEPWTBSKLPJNT5O4A/graph.json","events_json":"https://pith.science/api/pith-number/UPSKCNFDPIEPWTBSKLPJNT5O4A/events.json","paper":"https://pith.science/paper/UPSKCNFD"},"agent_actions":{"view_html":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A","download_json":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A.json","view_paper":"https://pith.science/paper/UPSKCNFD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.00079&json=true","fetch_graph":"https://pith.science/api/pith-number/UPSKCNFDPIEPWTBSKLPJNT5O4A/graph.json","fetch_events":"https://pith.science/api/pith-number/UPSKCNFDPIEPWTBSKLPJNT5O4A/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A/action/storage_attestation","attest_author":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A/action/author_attestation","sign_citation":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A/action/citation_signature","submit_replication":"https://pith.science/pith/UPSKCNFDPIEPWTBSKLPJNT5O4A/action/replication_record"}},"created_at":"2026-07-05T12:04:12.860110+00:00","updated_at":"2026-07-05T12:04:12.860110+00:00"}