{"paper":{"title":"AlayaLaser: Efficient Index Layout and Search Strategy for Large-scale High-dimensional Vector Similarity Search","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"AlayaLaser shows that on-disk graph indexes for high-dimensional vectors can match or beat in-memory speed by fixing compute bottlenecks instead of chasing I/O savings.","cross_cats":["cs.IR"],"primary_cat":"cs.DB","authors_text":"Bo Tang, Gezi Li, Haotian Liu, Liang Huang, Long Xiang, Weijian Chen, Yangshen Deng","submitted_at":"2026-02-26T18:48:29Z","abstract_excerpt":"On-disk graph-based approximate nearest neighbor search (ANNS) is essential for large-scale, high-dimensional vector retrieval, yet its performance is widely recognized to be limited by the prohibitive I/O costs. Interestingly, we observed that the performance of on-disk graph-based index systems is compute-bound, not I/O-bound, with the rising of the vector data dimensionality (e.g., hundreds or thousands). This insight uncovers a significant optimization opportunity: existing on-disk graph-based index systems universally target I/O reduction and largely overlook computational overhead, which"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"AlayaLaser not only surpasses existing on-disk graph-based index systems but also matches or even exceeds the performance of in-memory index systems.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the performance bottleneck of existing on-disk graph-based systems is primarily compute-bound rather than I/O-bound once dimensionality reaches hundreds or thousands, and that the proposed SIMD-friendly layout plus heuristics will reliably translate this insight into measurable gains across datasets.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"AlayaLaser uses a SIMD-optimized on-disk graph layout plus caching and search strategies to outperform prior on-disk ANNS systems and match or exceed in-memory performance on large high-dimensional datasets.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"AlayaLaser shows that on-disk graph indexes for high-dimensional vectors can match or beat in-memory speed by fixing compute bottlenecks instead of chasing I/O savings.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"66d5b20224ed74510a39e3bdbde4405b20f46c7cc9da453bf7eea6de6bc8dc71"},"source":{"id":"2602.23342","kind":"arxiv","version":2},"verdict":{"id":"76efd490-bb14-43ce-9a50-4c2f00bb1c4d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T18:41:56.212217Z","strongest_claim":"AlayaLaser not only surpasses existing on-disk graph-based index systems but also matches or even exceeds the performance of in-memory index systems.","one_line_summary":"AlayaLaser uses a SIMD-optimized on-disk graph layout plus caching and search strategies to outperform prior on-disk ANNS systems and match or exceed in-memory performance on large high-dimensional datasets.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the performance bottleneck of existing on-disk graph-based systems is primarily compute-bound rather than I/O-bound once dimensionality reaches hundreds or thousands, and that the proposed SIMD-friendly layout plus heuristics will reliably translate this insight into measurable gains across datasets.","pith_extraction_headline":"AlayaLaser shows that on-disk graph indexes for high-dimensional vectors can match or beat in-memory speed by fixing compute bottlenecks instead of chasing I/O savings."},"references":{"count":58,"sample":[{"doi":"","year":2023,"title":"Cecilia Aguerrebere, Ishwar Bhati, Mark Hildebrand, Mariano Tepper, and Ted Willke. 2023. Similarity search in the blink of an eye with compressed indices. arXiv preprint arXiv:2304.04759(2023)","work_id":"dc14ca5e-4004-4560-a6df-a5e4faad08cb","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"Fabien André, Anne-Marie Kermarrec, and Nicolas Le Scouarnec. 2017. Acceler- ated nearest neighbor search with quick adc. InICMR. 159–166","work_id":"e44c49f8-482f-47fd-87f4-4a64f54dfc1d","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"Fabien André, Anne-Marie Kermarrec, and Nicolas Le Scouarnec. 2019. Quicker adc: Unlocking the hidden potential of product quantization with simd.TPAMI 43, 5 (2019), 1666–1677","work_id":"6947de8e-5999-48be-a704-842186fd91fd","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Argilla. 2024. PersonaHub-FineWeb-Edu-4-Embeddings. Hugging Face dataset.https://huggingface.co/datasets/argilla-warehouse/personahub- fineweb-edu-4-embeddingsAccessed: 2025-10-06","work_id":"c109a04b-060b-4a6a-9814-9d333587c68f","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Martin Aumüller, Erik Bernhardsson, and Alexander Faithfull. 2020. ANN- Benchmarks: A benchmarking tool for approximate nearest neighbor algorithms. Information Systems87 (2020), 101374","work_id":"63a4a9b9-0e11-4e11-a0cc-e335ac7e45ae","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":58,"snapshot_sha256":"876ae6af4c57566c0272993f13f44d5517d53eb37f60f144c3773e8998a4b83a","internal_anchors":2},"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"}