CS Paper Radar

CoWalk: Cooperative Walks for Agentic Batch Queries on Disk-Resident Graphs

Qiao Liang, Zhiguang Chen, Yue Yu, Nong Xiao

cs.DB cs.IR ·submitted 2026-10-28 07:49 UTC · v1

Abstract

Workloads such as query rewriting, deep research and group-relative policy optimization (GRPO) issue batches of semantically related vector queries. Existing disk-resident approximate nearest neighbor (ANN) systems process these queries independently. On three workloads, 31 to 67 percent of the I/Os of a batch fetch a page that another query of the batch has already fetched. A page cache shared by the batch removes these I/Os, yet it makes a single-thread search only $1.1$ to $1.2{\times}$ faster, because the search time goes to distance computation and the cache does not change the walk of any query. We present CoWalk, which runs the queries of a batch as cooperating walks on one thread. Every query keeps its own candidate list, a page is read once per batch, and the full-precision vector on every fetched page is ranked by every query of the batch. A query therefore reaches the target recall with a shorter candidate list, which computes fewer distances and issues fewer I/Os. At recall@10 of 0.95, CoWalk answers a batch on one thread $1.6$ to $1.9{\times}$ faster than the fastest per-query baseline and issues $2.8$ to $6.9{\times}$ fewer I/Os. On a server it serves $3.0$ to $6.9{\times}$ the throughput of that baseline at a similar batch latency.

Closest papers

Nearest neighbours by meaning across the whole index.

PrairieFire: Igniting High-Throughput Updates in Large-Scale Approximate Nearest Neighbor Search

cs.DB ·2026-10-16 ·Yuchen Huang, Baiteng Ma, Yiping Sun, Xiao Chen, Rui Ma +7 more
similarity 0.69
Abstract

Arrays of modern NVMe SSDs are approaching memory in terms of bandwidth while offering much lower cost per gigabyte. Hence, approximate nearest neighbor search (ANNS) for billion-scale vectors in applications such as recommendation, search, and retrieval-augmented generation increasingly runs on SSD-based systems. Unfortunately, due to the limited update throughput of existing SSD-based solutions, the indexed corpus cannot keep pace with continuous modifications to data and embedding models, thus hindering broader deployment across more scenarios and larger scales. The bottleneck lies in CPU-intensive update paths, where foreground insertions incur frequent SSD accesses and background reassignments repeatedly relocate each vector, making high update throughput difficult to achieve even by exhausting available hardware. In this paper, using buffered updates and batched local rebuilds, we develop PrairieFire with cascaded adjustments, concurrency control, and CPU-GPU-SSD cooperation. These co-designs avoid frequent SSD accesses in foreground updates and heavy background reassignments, achieving high update throughput while maintaining excellent search performance. Evaluations show that, compared with state-of-the-art systems, PrairieFire can sustain approximately 3-35x update throughput while maintaining approximately 3-5x search throughput simultaneously. It also achieves approximately 1.5-8.7x cost efficiency. PrairieFire has been deployed in RedNote's production environments for two months, improving business metrics by higher freshness without frequent index rebuilds.

0.22

ConStore: Search-Overlapped Prefetch and Co-Retrieval-Aware Content Storage for Disk-Based ANNS

cs.DB ·2026-10-15 ·Zhutao Zhuang, Nong Xiao, Zhiguang Chen, Yue Yu
similarity 0.65
Abstract

Disk-based approximate nearest-neighbor search (ANNS) supports retrieval-augmented generation (RAG), search, recommendation, and advertising over corpora that exceed memory capacity. Existing ANNS research often focuses on vector search at small top-K, whereas industrial-scale applications require larger retrieval sets. Fetching the corresponding content from storage remains comparatively underexplored. Our profiling across million- to billion-vector corpora shows that, at large K, content fetch latency can match or exceed vector-search latency and cannot be ignored in end-to-end retrieval. We present ConStore, a content storage engine combining two complementary techniques to reduce end-to-end retrieval latency. Speculative Inline Content Prefetch (SICP) commonly abstracts graph-based and cluster-based ANNS as progressive SSD I/O that refines an in-flight candidate queue. It prefetches content for stable leading candidates under a bounded I/O budget, overlapping content reads with search without changing the final top-K results. Co-Retrieval-Pattern-Aware Grouping (RPAG) uses query logs to place frequently co-retrieved documents in contiguous segments. Workload-Adaptive Segment Reorganization (WASR) adapts this layout to workload shifts by selectively regrouping segments under a bounded rewrite budget. Group Content Fetch (GCF) exploits this layout to coalesce per-document random reads into batched, page-aware I/O requests. SICP works independently of the content layout, while RPAG and GCF reduce fetch cost even without prefetching. Across three multi-hop QA workloads with K from 10 to 1600, ConStore reduces end-to-end retrieval latency over the best per-document-read baseline by up to 48% on DiskANN and 25% on SPANN. Separate experiments on 100M- and billion-vector indexes demonstrate SICP's applicability at larger scales.

0.17

Aker: Density-Aware Approximate Caching for Vector Search (Extended Version)

cs.DB ·2026-09-03 ·Sukjoon Oh, Minki Kang, Dohyun Kim, Baotong Lu, Jing Liu +3 more
top venue similarity 0.64
Abstract · Extended version of the paper published in Proceedings of the VLDB Endowment (PVLDB), Vol. 19, No. 10, pp. 2727-2740, 2026

Disk-based approximate nearest neighbor search (ANNS) incurs high I/O overhead due to frequent disk accesses during index traversal. Approximate caching, which reuses the results of past queries to serve future similar queries, offers a promising approach to bypass expensive disk searches. However, existing approaches suffer from two key limitations. First, their approximate hit predicates fail to simultaneously achieve high throughput and high accuracy, as they do not adapt to the varying local neighbor density in high-dimensional spaces. Second, they lack an effective refresh mechanism to maintain cache correctness under vector updates. We present Aker, an approximate cache for disk-based ANNS. Aker addresses these limitations through two core design choices. First, we introduce a per-query similarity threshold, where each cache entry maintains its own threshold that is dynamically adjusted based on observed cache hit patterns. This design enables Aker to adapt to neighborhood densities to preserve both efficiency and accuracy. Second, we propose del-consistency, a consistency model for ANNS caches that applies deletions eagerly and insertions lazily. Under this model, Aker implements a low-overhead refresh mechanism that bounds cache staleness and preserves high search accuracy. We integrate Aker into pgvector and evaluate it on representative workloads. Aker improves recall by up to 64 percentage points over prior solutions and increases QPS by up to 3.2x, while using 0.6x the memory of pgvector's shared buffers.

0.68

BOA: Beamwidth Online Adaptation for Filtered-ANNS on a GPU

cs.DC ·2026-09-14 ·Farhana Akter Tumpa, Rajiv Gupta
similarity 0.64
Abstract

Filtered approximate nearest neighbor search, i.e. returning the top-k vectors nearest to a query vector among those satisfying one or more attribute predicates, has become a fundamental operation in modern vector search systems. Graph-based solutions employ beam search to solve a batch of queries in parallel for high throughput and employ high fixed beamwidth of 100 or greater for ensuring high recall. We observe that, given a batch of queries, more than half of the queries across multiple data sets can be solved precisely with a beamwidth of just 50 or less. Therefore, existing systems based on fixed high beamwidth sacrifice throughput to achieve high recall by forcing every query to search as thoroughly as the hardest query in the batch even though majority of queries can be resolved by a shallow search. In this paper we present a filtered ANNS engine for a single GPU named BOA that uses online beamwidth adaptation to customize the search effort across queries within a batch under multi-attribute range filters. We address the recall throughput tradeoff with a multi-phase search: all queries are first evaluated under a narrow beam, and only those with uncertain results are progressively refined with wider beamwidths. This renders recall largely insensitive to the starting beamwidth, whereas prior methods must use a fixed high beamwidth for high recall. BOA+ overlaps execution of phases to further enhance throughput. Our experiments show that, for 10,000 queries, online adaptation achieves 94.05% to 99.96% recall with average beamwidth ranging from 22 to 77, while a non-adaptive approach requires a fixed beamwidth of 500 to achieve similar or lower recall. Consequently, adaptivity increases throughput by 7x to 12.5x

0.07

Geo-Filtered Approximate Nearest-Neighbor Search on a Single Proximity Graph

cs.DB ·2026-10-25 ·Zhiqiu Zou, Rong-Hua Li, Qiangqiang Dai, Guoren Wang
theory similarity 0.64
Abstract · 15 pages, 9 figures, 5 tables. Full version with appendix and proofs

Geo-filtered approximate nearest-neighbor (GFANN) search retrieves the objects inside a query-specified geographic disk whose feature vectors are closest to a query vector, a core operation in location-aware Web search. A disk filter on a standard proximity graph invalidates out-of-disk routing vertices; on selective disks, these removals break greedy paths and recall drops sharply. Existing methods scan all in-disk objects, over-fetch unfiltered candidates, or maintain separate sub-indexes across spatial partitions, trading latency for index overhead. We propose REACH, a single-graph index based on row-local active-order invariance: if every vertex stores its outgoing neighbors in a fixed-width row sorted by vector distance, any disk filter masks out-of-disk entries while surviving in-disk (active) neighbors remain vector-sorted. Offline, each vertex merges globally closest vector candidates with a nearest-first geographic stream into one row; online, beam search evaluates vector distances only for the first l active entries of each visited row. Because co-located objects share the same distance to every query center, every disk admits a coordinate group in full or not at all, decomposing active connectivity into intra-group and inter-coordinate conditions. Under a spatial marked-site model with repeated coordinates, we prove that polylogarithmic out-degree guarantees strong connectivity across all disks and constant-factor distance reduction per hop on compact disks, while a cut-rank threshold characterizes the minimum scan budget l. Across four datasets up to 10^7 objects, one REACH index per dataset outperforms SOTA methods, achieving 2.49x higher geometric-mean QPS than the strongest baseline at 0.95 Recall@10 on selective queries, while building 3.8x-11.9x faster and using 37%-45% less space than a SOTA spatial-graph index.

0.27

HyperPage: Learned Page Layouts for Disk-Resident Graph Vector Indexes

cs.DB cs.IR ·2026-11-12 ·Xinyu Zhang, Jingze Li, Baihua Zheng, Weiwei Sun
theorycode/data similarity 0.64
Abstract · 14 pages, 4 figures, 7 tables. Code available at: https://github.com/zhangshea/HyperPage

Disk-resident graph indexes such as DiskANN store each vector with its graph edges as one record on SSD pages. A search needs the page of every record it expands, so its page reads depend on the page layout, that is, which records share a page. Search agents trained with reinforcement learning query one such index run after run, and every graph traversal can be logged with the records it expands. We therefore use the logs of a finished run to lay out the index for a declared memory budget, so later runs read fewer pages. In contrast, existing layouts such as Starling and MARGO are computed from the graph and vectors and use no logs. Learning a layout from logs is challenging, because the logs must stay valid under every layout, the saving depends on what the budget caches, and the best layout is NP-hard to find. Our key observation is that DiskANN-style search routes on in-memory codes and the records it expands, never on their page-mates. As a result, the logged traversals price every layout by replay, and cached records add no page reads. Accordingly, we propose HyperPage, which first builds layouts for the declared budget on the logged traversals of uncached records. It then selects one layout, together with the budget split between hot records and a page cache, by replaying validation steps. In addition, a Lagrangian lower bound shows how far a layout may be from the optimum. We learn from one Search-R1 agent run and test on a later run with a different reward, over 1M and 21M passages. Across seven cache budgets, HyperPage reads 5.2% to 21.2% fewer pages than MARGO, and no compared layout reads fewer. The advantage is largest when the cache is small. For example, at 1% of the 1M index, MARGO needs nearly 3 times the budget to read as few pages. On the 21M index, HyperPage also cuts search time per step by up to 10.4%. The saving holds for a second agent model and a second corpus.

0.47

Fidelity Synopses: Pre-Execution Plan Selection for Approximate Vector Search

cs.DB cs.IR ·2026-10-30 ·Adithya Ramanatha, Sanskar S. Singh, Maya Ramanath
benchmark similarity 0.64
Abstract

Vector search has become a database operator with many physical implementations: different index families, build-time and search-time parameters, and compressed or disk-resident variants. Each trades cost against fidelity, the fraction of a query's exact top-k nearest neighbors that a configuration returns. Deployments nevertheless choose one configuration offline and use it for every query, so they are sized for their hardest queries and overpay for the rest. On a 5.8M-item LAION benchmark with four embedding models, running each query on the cheapest configuration that suffices for it would roughly halve the cost in a typical setting. Exploiting this requires predicting each configuration's fidelity before running it, which is hard because fidelity is defined by the very exact search that approximate indexes avoid. Relational optimizers solve the analogous problem for cardinality with offline synopses; we propose one for vector search. A fidelity synopsis divides the embedding space into cells around a set of landmarks and records, for every cell and configuration, how often that configuration met a fidelity target on a sample of queries. At query time, one nearest-landmark lookup yields a calibrated estimate for every configuration, and an optimizer chooses configurations so that the workload misses the target no more often than its requirement allows. We test the two premises the synopsis relies on, that fidelity varies smoothly across the embedding space and that its loss is specific to each configuration. Held to the same requirement a static deployment is built to meet, synopsis-guided selection lowers cost in all 16 combinations of model and query direction, to a median of 0.72x the static deployment, with a batched lookup of 21-32 microseconds per query. The saving persists against the best static choice among the same indexes, and within a single index at no extra memory.

0.09

RiftANN: Efficient Graph Traversal for Vector Search with RDMA-Based Memory Disaggregation

cs.DB ·2026-10-06 ·Qi Lin, Zhenyu Zhang, Jun Kong, Zhichao Cao
similarity 0.63
Abstract

Graph indexes for billion-scale vector collections can exceed a single server's DRAM capacity. Passive RDMA-based disaggregated memory provides scalable capacity without memory-node computation, but conventional best-first search performs poorly in this setting. Fine-grained and unnecessary remote reads increase communication cost, while dependencies between candidate evaluation and subsequent expansion serialize traversal and leave CPUs idle. We present RiftANN, a graph-based approximate nearest neighbor search (ANNS) system for passive RDMA-based disaggregated memory. RiftANN navigates using compact vectors at compute nodes and selectively verifies exact vectors using a threshold calibrated from observed approximation errors. A bounded one-sided RDMA pipeline batches and overlaps the remaining remote accesses with local processing. RiftANN evaluates returned neighbors in parallel and incrementally incorporates completed results. An impact gate estimates whether unfinished evaluations may change the candidate set, allowing traversal to continue when their expected influence is small while preventing unsafe advancement from incomplete search state. A lightweight feedback controller coordinates RDMA concurrency with background evaluation capacity as search configurations and query loads change. We evaluate RiftANN on SIFT, DEEP, and SPACEV at 100M scale and SIFT1B. At matched recall, RiftANN achieves 1.6x to 4.3x latency speedups and 1.1x to 2.2x higher throughput than DistVS, and 1.6x to 5.6x latency speedups over SSD-based DiskANN and PipeANN. These results show that passive disaggregated memory can support low-latency graph-based vector retrieval without query-time computation at memory nodes.

0.21