{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-08-20T21:21:00.605Z","headline":"AlloyDB ScaNN 如何将向量搜索扩展到 100 亿向量","description":"AlloyDB 的 ScaNN 索引现已支持超过 100 亿向量的规模，通过全新的四层树架构（预览版）实现，将查询复杂度从 O（N^1/2） 降至 O（N^1/4）。内部测试中，该架构在 100 亿向量规模下可实现 p95 延迟不超过 51 毫秒、召回率达 95%。该功能可通过快速入门指南部署，新用户可享受 30 天免费试用。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","url":"https://www.aioga.com/news/cmt1r7jjh05lgroovc4zwypj6/","mainEntityOfPage":"https://www.aioga.com/news/cmt1r7jjh05lgroovc4zwypj6/","datePublished":"2026-08-20T16:00:00.000Z","dateModified":"2026-08-20T16:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://cloud.google.com/blog/products/databases/alloydb-scann-index-four-level-tree-improves-vector-search","https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6"],"canonicalUrl":"https://www.aioga.com/news/cmt1r7jjh05lgroovc4zwypj6/","directAnswer":{"@type":"Answer","text":"Google Cloud 表示，AlloyDB 的 ScaNN 索引通过处于预览阶段的四层树架构，可在 100 亿向量规模下高效运行。材料称，该架构旨在改善大规模向量检索中的计算与内存压力。","url":"https://www.aioga.com/news/cmt1r7jjh05lgroovc4zwypj6/","dateCreated":"2026-08-20T16:00:00.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"cloud.google.com source article","url":"https://cloud.google.com/blog/products/databases/alloydb-scann-index-four-level-tree-improves-vector-search","datePublished":"2026-08-20T16:00:00.000Z","provider":{"@type":"Organization","name":"cloud.google.com","url":"https://cloud.google.com/blog/products/databases/alloydb-scann-index-four-level-tree-improves-vector-search"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","datePublished":"2026-08-20T16:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6"}}],"aggregationSource":"Google Cloud：Databases（RSS","originalPublisher":{"name":"cloud.google.com","url":"https://cloud.google.com/blog/products/databases/alloydb-scann-index-four-level-tree-improves-vector-search"},"geoDeepAnswer":null,"article":{"id":"cmt1r7jjh05lgroovc4zwypj6","slug":"cmt1r7jjh05lgroovc4zwypj6","url":"https://www.aioga.com/news/cmt1r7jjh05lgroovc4zwypj6/","title":"AlloyDB ScaNN 如何将向量搜索扩展到 100 亿向量","title_en":"","summary":"AlloyDB 的 ScaNN 索引现已支持超过 100 亿向量的规模，通过全新的四层树架构（预览版）实现，将查询复杂度从 O（N^1/2） 降至 O（N^1/4）。内部测试中，该架构在 100 亿向量规模下可实现 p95 延迟不超过 51 毫秒、召回率达 95%。该功能可通过快速入门指南部署，新用户可享受 30 天免费试用。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","source":"Google Cloud：Databases（RSS","sourceUrl":"https://cloud.google.com/blog/products/databases/alloydb-scann-index-four-level-tree-improves-vector-search","aiHotUrl":"https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","publishedAt":"2026-08-20T16:00:00.000Z","category":"行业动态","score":72,"selected":true,"articleBody":["The front door to AI in the workplace","To satisfy the demands of enterprise-grade agentic AI applications, underlying vector databases often struggle to scale effectively as modern use cases can scale to billions of vectors.","As a fully managed PostgreSQL-compatible database service, AlloyDB ：https://docs.cloud.google.com/alloydb/docs/overview is engineered to handle demanding enterprise workloads. Combining Google's infrastructure with the reliability of commercial databases, it delivers high availability, scalability, and includes a cutting-edge analytical engine, optimal for agentic AI use cases. A key part of this is its ScaNN index ：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index , which now operates efficiently at a scale of 10 billion vectors . This was achieved through a major architectural enhancement: an innovative four-level tree (preview) ：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index paired with efficient memory usage .","Scaling to a 10 billion vector workload presents significant memory and computational challenges. Previous AlloyDB ScaNN tree-based index was limited to two ：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#two-level-tree-index - or three ：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#three-level-tree-index -level tree configurations, and attempting to scale those structures led to several bottlenecks:","Increased compute intensity: Larger tree structures demand significantly more operations for both index construction and query traversal.","Memory constraints: The sampling processes required for 10 billion vectors can easily exceed the system's available memory capacity.","The introduction of a four-level tree (preview) ：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index is the primary innovation in the recent AlloyDB ScaNN release. This architecture, illustrated in Figure 1, employs a top-down strategy to optimize the balance between accuracy and build efficiency. To maintain high performance and mitigate recall loss, the system integrates key enhancements such as Top-K branch, SOAR ：https://research.google/blog/soar-new-algorithms-for-even-faster-vector-search-with-scann/#:~:text=ScaNN%20is%20open%2Dsourced%20on%20GitHub%20and%20can%20be%20easily%20installed%20via%20Pip. , centroid adjustment ：https://arxiv.org/abs/1908.10396 and balanced tree shape.","Figure 1. AlloyDB ScaNN four-level tree architecture","This design has two primary benefits:","1. Reduced compute intensity via hierarchical partitioning","The four-level architecture drastically reduces compute intensity by using hierarchical partitioning to restrict the volume of vectors scanned during a query. Instead of traversing a flat or poorly segmented space, the multi-layered hierarchy narrows down the search path exponentially. Figure 2 illustrates the search spaces across different tree levels, demonstrating how structural layering optimizes traversal efficiency:","Figure 2. Search space for two-, three- and four-level trees","Two-level: Utilizes coarse partitioning to guide queries, resulting in a basic search complexity of O( N 1/2 ) .","Three-level: Introduces an intermediate layer to further subdivide clusters, narrowing exploration to O( N 1/3 ) .","Four-level: Implements refined, highly granular partitions that optimize traversal efficiency down to O( N 1/4 ) , sufficiently allowing for more than 10-billion vectors.","By dynamically expanding hierarchical layers as the dataset expands, AlloyDB ScaNN maintains ultra-low query latency and avoids computational scale walls from impacting performance.","Achieving a 10 billion vector scale requires high memory efficiency. AlloyDB ScaNN uses these strategies to maximize memory management performance:","Balanced tree shape construction: The four-level tree utilizes a balanced configuration to circumvent memory limitations that restrict the size of training datasets. This balanced architecture effectively leverages reduced sampling sizes to construct high-fidelity tree partitions.","Sampling optimization: When the system encounters memory limitations, it generates a condensed sampling set that considers performance and accuracy.","By leveraging the innovative four-level tree architecture in our internal tests, we are able to achieve the following performance results:","AlloyDB can scale to over 10 billion vectors with its ScaNN index.","Experience AlloyDB ScaNN's four-level tree (preview) ：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index architecture today. You can deploy ScaNN for AlloyDB by following our quickstart guide ：https://cloud.google.com/alloydb/docs/quickstart/create-and-connect to set up an instance. For optimized, high-speed vector search, refer to the official ScaNN documentation ：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index . New users can also explore AlloyDB ：https://console.cloud.google.com/alloydb/create-trial-cluster?_gl=1*qsd2cd*_up*MQ..&gclid=CjwKCAjwooq3BhB3EiwAYqYoEh91xxGzv4xrmyMJJ_BPfF4X8cv-I3kINwvnMI2pADozFQPsrHnaOhoCbioQAvD_BwE&gclsrc=aw.ds through our 30-day free trial ：https://cloud.google.com/blog/products/databases/run-your-postgresql-database-in-an-alloydb-free-trial-cluster program. We can’t wait to hear about what you build!","By Niranjan Shivprasad • 6-minute read"],"articleImages":[{"sourceUrl":"https://storage.googleapis.com/gweb-cloudblog-publish/images/1_fpfICUj.max-1000x1000.jpg","alt":"https://storage.googleapis.com/gweb-cloudblog-publish/images/1_fpfICUj.max-1000x1000.jpg","afterParagraph":6,"url":"/media/articles/cmt1r7jjh05lgroovc4zwypj6/27f2ed8834ed75fc.jpg"},{"sourceUrl":"https://storage.googleapis.com/gweb-cloudblog-publish/images/2_LWwXC70.max-1000x1000.jpg","alt":"https://storage.googleapis.com/gweb-cloudblog-publish/images/2_LWwXC70.max-1000x1000.jpg","afterParagraph":10,"url":"/media/articles/cmt1r7jjh05lgroovc4zwypj6/6753b559b62d868f.jpg"},{"sourceUrl":"https://storage.googleapis.com/gweb-cloudblog-publish/images/09_-_Data_Analytics_tFH57V6.max-700x700.jpg","alt":"https://storage.googleapis.com/gweb-cloudblog-publish/images/09_-_Data_Analytics_tFH57V6.max-700x700.jpg","afterParagraph":21,"url":"/media/articles/cmt1r7jjh05lgroovc4zwypj6/acea37812db8fda3.jpg"},{"sourceUrl":"https://storage.googleapis.com/gweb-cloudblog-publish/images/Hero_Image_KjCoerO.max-700x700.jpg","alt":"https://storage.googleapis.com/gweb-cloudblog-publish/images/Hero_Image_KjCoerO.max-700x700.jpg","afterParagraph":21,"url":"/media/articles/cmt1r7jjh05lgroovc4zwypj6/cb3abb0b0d4f7fc8.jpg"},{"sourceUrl":"https://storage.googleapis.com/gweb-cloudblog-publish/images/10_-_Databases.max-700x700.jpg","alt":"https://storage.googleapis.com/gweb-cloudblog-publish/images/10_-_Databases.max-700x700.jpg","afterParagraph":21,"url":"/media/articles/cmt1r7jjh05lgroovc4zwypj6/6ab573326db35ef3.jpg"},{"sourceUrl":"https://storage.googleapis.com/gweb-cloudblog-publish/images/29_-_Retail_HmMLc8R.max-700x700.jpg","alt":"https://storage.googleapis.com/gweb-cloudblog-publish/images/29_-_Retail_HmMLc8R.max-700x700.jpg","afterParagraph":22,"url":"/media/articles/cmt1r7jjh05lgroovc4zwypj6/ae3b6841fb418b0b.jpg"}],"mediaStatus":"ok","articleBodyZh":["通向职场人工智能的前门","为了满足企业级自主 AI 应用的需求，底层向量数据库通常在扩展方面难以高效应对，因为现代用例可能扩展到数十亿个向量。","作为一个完全托管的兼容 PostgreSQL 的数据库服务，AlloyDB：https://docs.cloud.google.com/alloydb/docs/overview 专为处理高需求企业工作负载而设计。它将 Google 的基础设施与商业数据库的可靠性结合，提供高可用性、可扩展性，并包含先进的分析引擎，非常适合自主 AI 用例。其中一个关键部分是其 ScaNN 索引：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index，现在可在 100 亿向量的规模下高效运行。这是通过一次重大的架构改进实现的：创新的四级树（预览）：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index 并配合高效的内存使用。","扩展到 100 亿向量的工作负载会带来显著的内存和计算挑战。此前的 AlloyDB ScaNN 基于树的索引仅限于二级：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#two-level-tree-index 或三级：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#three-level-tree-index 树结构，尝试扩展这些结构会导致多个瓶颈：","计算强度增加：更大的树结构在索引构建和查询遍历过程中需要更多操作。","内存限制：处理 100 亿向量所需的采样过程很容易超过系统可用的内存容量。","在最近的 AlloyDB ScaNN 版本中，四级树（预览版）的引入：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index 是主要创新。该架构如图 1 所示，采用自上而下的策略优化精度与构建效率之间的平衡。为了保持高性能并减轻召回损失，系统集成了关键增强功能，例如 Top-K 分支、SOAR：https://research.google/blog/soar-new-algorithms-for-even-faster-vector-search-with-scann/#:~:text=ScaNN%20is%20open%2Dsourced%20on%20GitHub%20and%20can%20be%20easily%20installed%20via%20Pip.，质心调整：https://arxiv.org/abs/1908.10396 以及平衡树形结构。","图 1. AlloyDB ScaNN 四级树架构","这一设计有两个主要优点：","1. 通过分层划分减少计算强度","四级架构通过使用分层划分在查询时限制扫描向量的数量，从而显著降低计算强度。与遍历平坦或分割不良的空间不同，多层级层次结构呈指数级缩小搜索路径。图 2 展示了不同树级别的搜索空间，说明结构分层如何优化遍历效率：","图 2. 二级、三级和四级树的搜索空间","二级：利用粗粒度划分引导查询，基本搜索复杂度为 O( N 1/2 )。","三级：引入中间层进一步细分簇群，将探索范围缩小至 O( N 1/3 )。","四级：实施精细、高粒度的分区，将遍历效率优化至 O( N 1/4 )，足以处理超过 100 亿个向量。","通过在数据集扩展时动态增加分层层数，AlloyDB ScaNN 保持超低查询延迟，并避免计算规模壁垒影响性能。","实现 100 亿向量规模需要高内存效率。AlloyDB ScaNN 使用以下策略最大化内存管理性能：","平衡树形结构构建：四层树利用平衡配置来规避限制训练数据集大小的内存限制。这种平衡的架构能有效利用减少的采样量来构建高保真度的树分区。","采样优化：当系统遇到内存限制时，会生成一个考虑性能和准确性的精简采样集。","通过在内部测试中利用创新的四层树架构，我们能够实现以下性能结果：","AlloyDB 可以使用其 ScaNN 索引扩展至超过 100 亿个向量。","立即体验 AlloyDB ScaNN 的四层树（预览版）：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index 架构。您可以通过我们的快速入门指南部署 AlloyDB 的 ScaNN：https://cloud.google.com/alloydb/docs/quickstart/create-and-connect 来设置实例。有关优化的高速向量搜索，请参阅官方 ScaNN 文档：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index。新用户还可以通过我们的 30 天免费试用计划探索 AlloyDB：https://console.cloud.google.com/alloydb/create-trial-cluster?_gl=1*qsd2cd*_up*MQ..&gclid=CjwKCAjwooq3BhB3EiwAYqYoEh91xxGzv4xrmyMJJ_BPfF4X8cv-I3kINwvnMI2pADozFQPsrHnaOhoCbioQAvD_BwE&gclsrc=aw.ds：https://cloud.google.com/blog/products/databases/run-your-postgresql-database-in-an-alloydb-free-trial-cluster。我们迫不及待想听听你们的成果！","作者：Niranjan Shivprasad • 阅读时间：6 分钟"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Google Cloud 表示，AlloyDB 的 ScaNN 索引通过处于预览阶段的四层树架构，可在 100 亿向量规模下高效运行。材料称，该架构旨在改善大规模向量检索中的计算与内存压力。","background":"正文指出，面向企业级智能体 AI 应用的向量数据库，可能需要处理数十亿级向量。此前 AlloyDB ScaNN 树索引仅提供两层或三层配置；扩展至 100 亿向量时，索引构建、查询遍历和采样内存都会形成瓶颈。","viewpoint":"Aioga 判断，这次更新的重点不只是提高可承载的向量数量，而是以更深的树结构重构大规模检索路径。摘要中的复杂度变化和内部测试指标值得关注，但均应视为厂商披露的预览版性能信息。","implications":"对需要在 PostgreSQL 兼容托管数据库中部署大规模向量搜索的团队而言，四层树提供了新的索引选择。材料称其结合 Top-K branch、SOAR、质心调整与平衡树形，以兼顾准确性和构建效率；实际收益可能取决于数据与查询负载。","nextStep":"计划评估该能力的团队，应先确认四层树仍处于预览阶段，并依据官方快速入门与索引配置文档建立代表性测试。重点对比召回率、p95 延迟、索引构建成本和内存占用，避免直接将内部测试结果外推至生产环境。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-08-20T17:03:19.328Z","sourceHash":"f38528b878d2596b","review":{"approved":true,"groundedness":94,"clarity":91,"duplicationRisk":12,"blockingIssues":[],"notes":["“Aioga 判断”已明确标注为观点；关于复杂度、p95 延迟和召回率的内容也注明为厂商披露的预览版性能信息，未冒充独立验证结果。","“实际收益可能取决于数据与查询负载”属于合理的条件性分析，不构成事实性断言。","可选措辞优化：可将“重构大规模检索路径”改为“优化大规模检索路径”，以减少“重构”可能带来的过度解读。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","Google Cloud：Databases（RSS）"],"translations":{"zh-CN":{"title":"AlloyDB ScaNN 如何将向量搜索扩展到 100 亿向量","summary":"AlloyDB 的 ScaNN 索引现已支持超过 100 亿向量的规模，通过全新的四层树架构（预览版）实现，将查询复杂度从 O（N^1/2） 降至 O（N^1/4）。内部测试中，该架构在 100 亿向量规模下可实现 p95 延迟不超过 51 毫秒、召回率达 95%。该功能可通过快速入门指南部署，新用户可享受 30 天免费试用。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"cloud.google.com","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"AlloyDB ScaNN 如何将向量搜索扩展到 100 亿向量 - Aioga AI资讯","description":"AlloyDB 的 ScaNN 索引现已支持超过 100 亿向量的规模，通过全新的四层树架构（预览版）实现，将查询复杂度从 O（N^1/2） 降至 O（N^1/4）。内部测试中，该架构在 100 亿向量规模下可实现 p95 延迟不超过 51 毫秒、召回率达 95%。该功能可通过快速入门指南部署，新用户可享受 30 天免费试用。 🔗 阅读原文 via AIHO...","url":"https://www.aioga.com/news/cmt1r7jjh05lgroovc4zwypj6/","articleBody":["通向职场人工智能的前门","为了满足企业级自主 AI 应用的需求，底层向量数据库通常在扩展方面难以高效应对，因为现代用例可能扩展到数十亿个向量。","作为一个完全托管的兼容 PostgreSQL 的数据库服务，AlloyDB：https://docs.cloud.google.com/alloydb/docs/overview 专为处理高需求企业工作负载而设计。它将 Google 的基础设施与商业数据库的可靠性结合，提供高可用性、可扩展性，并包含先进的分析引擎，非常适合自主 AI 用例。其中一个关键部分是其 ScaNN 索引：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index，现在可在 100 亿向量的规模下高效运行。这是通过一次重大的架构改进实现的：创新的四级树（预览）：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index 并配合高效的内存使用。","扩展到 100 亿向量的工作负载会带来显著的内存和计算挑战。此前的 AlloyDB ScaNN 基于树的索引仅限于二级：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#two-level-tree-index 或三级：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#three-level-tree-index 树结构，尝试扩展这些结构会导致多个瓶颈：","计算强度增加：更大的树结构在索引构建和查询遍历过程中需要更多操作。","内存限制：处理 100 亿向量所需的采样过程很容易超过系统可用的内存容量。","在最近的 AlloyDB ScaNN 版本中，四级树（预览版）的引入：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index 是主要创新。该架构如图 1 所示，采用自上而下的策略优化精度与构建效率之间的平衡。为了保持高性能并减轻召回损失，系统集成了关键增强功能，例如 Top-K 分支、SOAR：https://research.google/blog/soar-new-algorithms-for-even-faster-vector-search-with-scann/#:~:text=ScaNN%20is%20open%2Dsourced%20on%20GitHub%20and%20can%20be%20easily%20installed%20via%20Pip.，质心调整：https://arxiv.org/abs/1908.10396 以及平衡树形结构。","图 1. AlloyDB ScaNN 四级树架构","这一设计有两个主要优点：","1. 通过分层划分减少计算强度","四级架构通过使用分层划分在查询时限制扫描向量的数量，从而显著降低计算强度。与遍历平坦或分割不良的空间不同，多层级层次结构呈指数级缩小搜索路径。图 2 展示了不同树级别的搜索空间，说明结构分层如何优化遍历效率：","图 2. 二级、三级和四级树的搜索空间","二级：利用粗粒度划分引导查询，基本搜索复杂度为 O( N 1/2 )。","三级：引入中间层进一步细分簇群，将探索范围缩小至 O( N 1/3 )。","四级：实施精细、高粒度的分区，将遍历效率优化至 O( N 1/4 )，足以处理超过 100 亿个向量。","通过在数据集扩展时动态增加分层层数，AlloyDB ScaNN 保持超低查询延迟，并避免计算规模壁垒影响性能。","实现 100 亿向量规模需要高内存效率。AlloyDB ScaNN 使用以下策略最大化内存管理性能：","平衡树形结构构建：四层树利用平衡配置来规避限制训练数据集大小的内存限制。这种平衡的架构能有效利用减少的采样量来构建高保真度的树分区。","采样优化：当系统遇到内存限制时，会生成一个考虑性能和准确性的精简采样集。","通过在内部测试中利用创新的四层树架构，我们能够实现以下性能结果：","AlloyDB 可以使用其 ScaNN 索引扩展至超过 100 亿个向量。","立即体验 AlloyDB ScaNN 的四层树（预览版）：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index#four-level-tree-index 架构。您可以通过我们的快速入门指南部署 AlloyDB 的 ScaNN：https://cloud.google.com/alloydb/docs/quickstart/create-and-connect 来设置实例。有关优化的高速向量搜索，请参阅官方 ScaNN 文档：https://docs.cloud.google.com/alloydb/docs/ai/create-scann-index。新用户还可以通过我们的 30 天免费试用计划探索 AlloyDB：https://console.cloud.google.com/alloydb/create-trial-cluster?_gl=1*qsd2cd*_up*MQ..&gclid=CjwKCAjwooq3BhB3EiwAYqYoEh91xxGzv4xrmyMJJ_BPfF4X8cv-I3kINwvnMI2pADozFQPsrHnaOhoCbioQAvD_BwE&gclsrc=aw.ds：https://cloud.google.com/blog/products/databases/run-your-postgresql-database-in-an-alloydb-free-trial-cluster。我们迫不及待想听听你们的成果！","作者：Niranjan Shivprasad • 阅读时间：6 分钟"]},"en":{"title":"How AlloyDB ScaNN Scales Vector Search to 10 Billion Vectors","summary":"AlloyDB’s ScaNN index now supports scales of over 10 billion vectors through a brand-new four-layer tree architecture (preview), reducing query complexity from O(N^1/2) to O(N^1/4). In internal tests, this architecture achieved a p95 latency of no more than 51 ms and a recall rate of 95% at the 10 billion vector scale. This feature can be deployed via the Quick Start Guide, and new users can enjoy a 30-day free trial. 🔗 Read the original via AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"Industry","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"How AlloyDB ScaNN Scales Vector Search to 10 Billion Vectors - Aioga AI News","description":"AlloyDB’s ScaNN index now supports scales of over 10 billion vectors through a brand-new four-layer tree architecture (preview), reducing query complexity from O(N^1/2) to O(N^1/4)...","url":"https://www.aioga.com/en/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:27.090Z"},"ja":{"title":"AlloyDB ScaNNがベクター探索を100億ベクターに拡張する方法","summary":"AlloyDBのScaNNインデックスは、クエリの複雑さをO(N^1/2)からO(N^1/4)に削減する、新しい4層木構造(プレビュー)を通じて100億ベクトルを超えるスケールをサポートしました。 内部テストでは、このアーキテクチャは100億ベクトルスケールでP95レイテンシが51ミリ秒以下、リコール率95%を達成しています。 この機能はクイックスタートガイドを通じて展開でき、新規ユーザーは30日間の無料トライアルを利用できます。 🔗 原文記事はAIHOTより読むことができます。 https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"業界動向","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"AlloyDB ScaNNがベクター探索を100億ベクターに拡張する方法 - Aioga AIニュース","description":"AlloyDBのScaNNインデックスは、クエリの複雑さをO(N^1/2)からO(N^1/4)に削減する、新しい4層木構造(プレビュー)を通じて100億ベクトルを超えるスケールをサポートしました。 内部テストでは、このアーキテクチャは100億ベクトルスケールでP95レイテンシが51ミリ秒以下、リコール率95%を達成しています。 この機能はクイックスタートガイ...","url":"https://www.aioga.com/ja/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:28.118Z"},"ko":{"title":"AlloyDB ScaNN이 벡터 탐색을 100억 벡터로 확장하는 방법","summary":"AlloyDB의 ScaNN 인덱스는 이제 100억 벡터를 넘는 스케일을 지원하며, 쿼리 복잡도를 O(N^1/2)에서 O(N^1/4)로 줄인 새로운 4계층 트리 아키텍처(미리보기)를 통해 구현되었습니다. 내부 테스트에서 이 아키텍처는 100억 벡터 규모에서 P95 지연 시간(51밀리초)을 넘지 않고 95%의 회상률을 달성합니다. 이 기능은 빠른 시작 가이드를 통해 배포할 수 있으며, 신규 사용자는 30일 무료 체험을 누릴 수 있습니다. 🔗 원문 기사는 AIHOT를 통해 읽을 수 있습니다. https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"업계 동향","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"AlloyDB ScaNN이 벡터 탐색을 100억 벡터로 확장하는 방법 - Aioga AI 뉴스","description":"AlloyDB의 ScaNN 인덱스는 이제 100억 벡터를 넘는 스케일을 지원하며, 쿼리 복잡도를 O(N^1/2)에서 O(N^1/4)로 줄인 새로운 4계층 트리 아키텍처(미리보기)를 통해 구현되었습니다. 내부 테스트에서 이 아키텍처는 100억 벡터 규모에서 P95 지연 시간(51밀리초)을 넘지 않고 95%의 회상률을 달성합...","url":"https://www.aioga.com/ko/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:36.740Z"},"es":{"title":"Cómo AlloyDB ScaNN amplía la búsqueda vectorial a 10.000 millones de vectores","summary":"El índice ScaNN de AlloyDB ahora soporta escalas superiores a 10.000 millones de vectores, implementadas mediante una arquitectura de árbol de cuatro capas (preview) completamente nueva que reduce la complejidad de la consulta de O(N^1/2) a O(N^1/4). En pruebas internas, esta arquitectura alcanza una latencia P95 de no más de 51 milisegundos y una tasa de recuperación del 95% a una escala de 10.000 millones de vectores. Esta función puede desplegarse a través de la Guía de Inicio Rápido, y los nuevos usuarios pueden disfrutar de una prueba gratuita de 30 días. 🔗 Lee el artículo original a través de AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"Industria","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Cómo AlloyDB ScaNN amplía la búsqueda vectorial a 10.000 millones de vectores - Aioga Noticias de IA","description":"El índice ScaNN de AlloyDB ahora soporta escalas superiores a 10.000 millones de vectores, implementadas mediante una arquitectura de árbol de cuatro capas (preview) completamente...","url":"https://www.aioga.com/es/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:36.754Z"},"fr":{"title":"Comment AlloyDB ScaNN étend la recherche vectorielle à 10 milliards de vecteurs","summary":"L’indice ScaNN d’AlloyDB prend désormais en charge des échelles dépassant 10 milliards de vecteurs, implémentées via une toute nouvelle architecture d’arbre à quatre couches (preview) qui réduit la complexité des requêtes de O(N^1/2) à O(N^1/4). Lors des tests internes, cette architecture atteint une latence P95 ne dépassant pas 51 millisecondes et un taux de rappel de 95 % à une échelle de 10 milliards de vecteurs. Cette fonctionnalité peut être déployée via le Guide de démarrage rapide, et les nouveaux utilisateurs peuvent profiter d’un essai gratuit de 30 jours. 🔗 Lisez l’article original via AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"Industrie","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Comment AlloyDB ScaNN étend la recherche vectorielle à 10 milliards de vecteurs - Aioga Actualités IA","description":"L’indice ScaNN d’AlloyDB prend désormais en charge des échelles dépassant 10 milliards de vecteurs, implémentées via une toute nouvelle architecture d’arbre à quatre couches (previ...","url":"https://www.aioga.com/fr/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:45.265Z"},"de":{"title":"Wie AlloyDB ScaNN die Vektorsuche auf 10 Milliarden Vektoren erweitert","summary":"Der ScaNN-Index von AlloyDB unterstützt nun Skalierungen von über 10 Milliarden Vektoren, implementiert durch eine brandneue vierschichtige Baumarchitektur (Vorschau), die die Abfragekomplexität von O(N^1/2) auf O(N^1/4) reduziert. Bei internen Tests erreicht diese Architektur eine P95-Latenz von nicht mehr als 51 Millisekunden und eine Rückrufrate von 95 % auf einer Skala von 10 Milliarden Vektoren. Diese Funktion kann über den Quick Start Guide bereitgestellt werden, und neue Nutzer können eine 30-tägige kostenlose Testphase genießen. 🔗 Lesen Sie den Originalartikel über AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Wie AlloyDB ScaNN die Vektorsuche auf 10 Milliarden Vektoren erweitert - Aioga KI-News","description":"Der ScaNN-Index von AlloyDB unterstützt nun Skalierungen von über 10 Milliarden Vektoren, implementiert durch eine brandneue vierschichtige Baumarchitektur (Vorschau), die die Abfr...","url":"https://www.aioga.com/de/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:44.789Z"},"pt-BR":{"title":"Como o AlloyDB ScaNN Estende a Busca Vetorial para 10 Bilhões de Vetores","summary":"O índice ScaNN do AlloyDB agora suporta escalas superiores a 10 bilhões de vetores, implementadas por meio de uma nova arquitetura de árvore de quatro camadas (prévia) que reduz a complexidade da consulta de O(N^1/2) para O(N^1/4). Em testes internos, essa arquitetura alcança uma latência P95 de no máximo 51 milissegundos e uma taxa de recall de 95% em uma escala de 10 bilhões de vetores. Esse recurso pode ser implementado através do Quick Start Guide, e novos usuários podem desfrutar de um teste gratuito de 30 dias. 🔗 Leia o artigo original via AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Como o AlloyDB ScaNN Estende a Busca Vetorial para 10 Bilhões de Vetores - Aioga Notícias de IA","description":"O índice ScaNN do AlloyDB agora suporta escalas superiores a 10 bilhões de vetores, implementadas por meio de uma nova arquitetura de árvore de quatro camadas (prévia) que reduz a...","url":"https://www.aioga.com/pt-BR/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:53.939Z"},"ru":{"title":"Как AlloyDB ScaNN расширяет векторный поиск до 10 миллиардов векторов","summary":"Индекс ScaNN от AlloyDB теперь поддерживает масштабы, превышающие 10 миллиардов векторов, реализованные через совершенно новую архитектуру четырёхслойного дерева (предварительный просмотр), которая снижает сложность запросов с O(N^1/2) до O(N^1/4). Во внутреннем тестировании эта архитектура достигает задержки P95 не более 51 миллисекунды и частоты отзыва 95% при масштабе 10 миллиардов векторов. Эту функцию можно развернуть через Руководство быстрого запуска, и новые пользователи смогут получить 30-дневный бесплатный пробный период. 🔗 Прочитайте оригинальную статью на сайте AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Как AlloyDB ScaNN расширяет векторный поиск до 10 миллиардов векторов - Aioga Новости ИИ","description":"Индекс ScaNN от AlloyDB теперь поддерживает масштабы, превышающие 10 миллиардов векторов, реализованные через совершенно новую архитектуру четырёхслойного дерева (предварительный п...","url":"https://www.aioga.com/ru/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:01:53.807Z"},"ar":{"title":"كيف يوسع AlloyDB ScaNN البحث المتجه ليصل إلى 10 مليارات متجه","summary":"يدعم مؤشر ScaNN من AlloyDB الآن مقاييس تتجاوز 10 مليارات متجه، يتم تنفيذها من خلال بنية شجرة جديدة مكونة من أربع طبقات (معاينة) تقلل تعقيد الاستعلام من O(N^1/2) إلى O(N^1/4). في الاختبارات الداخلية، تحقق هذه البنية زمن استجابة P95 لا يزيد عن 51 مللي ثانية ومعدل استدعاء 95٪ على نطاق 10 مليارات متجه. يمكن نشر هذه الميزة من خلال دليل البدء السريع، ويمكن للمستخدمين الجدد الاستمتاع بتجربة مجانية لمدة 30 يوما. 🔗 اقرأ المقال الأصلي عبر AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"كيف يوسع AlloyDB ScaNN البحث المتجه ليصل إلى 10 مليارات متجه - Aioga أخبار الذكاء الاصطناعي","description":"يدعم مؤشر ScaNN من AlloyDB الآن مقاييس تتجاوز 10 مليارات متجه، يتم تنفيذها من خلال بنية شجرة جديدة مكونة من أربع طبقات (معاينة) تقلل تعقيد الاستعلام من O(N^1/2) إلى O(N^1/4). في ال...","url":"https://www.aioga.com/ar/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:00.978Z"},"hi":{"title":"कैसे AlloyDB ScaNN वेक्टर खोज को 10 बिलियन वैक्टर तक बढ़ाता है","summary":"AlloyDB का ScaNN इंडेक्स अब 10 बिलियन से अधिक वैक्टर से अधिक के स्केल का समर्थन करता है, जिसे एक नए चार-परत ट्री आर्किटेक्चर (पूर्वावलोकन) के माध्यम से लागू किया गया है जो क्वेरी जटिलता को O(N^1/2) से O(N^1/4) तक कम करता है। आंतरिक परीक्षण में, यह आर्किटेक्चर 95 मिलीसेकंड से अधिक की P51 विलंबता और 95 बिलियन वैक्टर के पैमाने पर 10% की रिकॉल दर प्राप्त करता है। इस सुविधा को क्विक स्टार्ट गाइड के माध्यम से तैनात किया जा सकता है, और नए उपयोगकर्ता 30-दिन के निःशुल्क परीक्षण का आनंद ले सकते हैं। 🔗 AIHOT के माध्यम से मूल लेख पढ़ें · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"कैसे AlloyDB ScaNN वेक्टर खोज को 10 बिलियन वैक्टर तक बढ़ाता है - Aioga AI समाचार","description":"AlloyDB का ScaNN इंडेक्स अब 10 बिलियन से अधिक वैक्टर से अधिक के स्केल का समर्थन करता है, जिसे एक नए चार-परत ट्री आर्किटेक्चर (पूर्वावलोकन) के माध्यम से लागू किया गया है जो क्वेरी ज...","url":"https://www.aioga.com/hi/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:01.881Z"},"it":{"title":"Come AlloyDB ScaNN estende la ricerca vettoriale a 10 miliardi di vettori","summary":"L'indice ScaNN di AlloyDB ora supporta scale superiori a 10 miliardi di vettori, implementate tramite una nuovissima architettura ad albero a quattro livelli (preview) che riduce la complessità delle query da O(N^1/2) a O(N^1/4). Nei test interni, questa architettura raggiunge una latenza P95 non superiore a 51 millisecondi e un tasso di richiamo del 95% su una scala di 10 miliardi di vettori. Questa funzione può essere implementata tramite la Guida Quick Start e i nuovi utenti possono godere di una prova gratuita di 30 giorni. 🔗 Leggi l'articolo originale su AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Come AlloyDB ScaNN estende la ricerca vettoriale a 10 miliardi di vettori - Aioga Notizie IA","description":"L'indice ScaNN di AlloyDB ora supporta scale superiori a 10 miliardi di vettori, implementate tramite una nuovissima architettura ad albero a quattro livelli (preview) che riduce l...","url":"https://www.aioga.com/it/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:10.410Z"},"nl":{"title":"Hoe AlloyDB ScaNN vectorzoekopdrachten uitbreidt tot 10 miljard vectoren","summary":"De ScaNN-index van AlloyDB ondersteunt nu schalen van meer dan 10 miljard vectoren, geïmplementeerd via een gloednieuwe vierlaagse boomarchitectuur (preview) die de querycomplexiteit vermindert van O(N^1/2) naar O(N^1/4). Bij interne tests bereikt deze architectuur een P95-latentie van niet meer dan 51 milliseconden en een terugroeppercentage van 95% op een schaal van 10 miljard vectoren. Deze functie kan worden uitgerold via de Quick Start Guide, en nieuwe gebruikers kunnen genieten van een gratis proefperiode van 30 dagen. 🔗 Lees het originele artikel via AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Hoe AlloyDB ScaNN vectorzoekopdrachten uitbreidt tot 10 miljard vectoren - Aioga AI-nieuws","description":"De ScaNN-index van AlloyDB ondersteunt nu schalen van meer dan 10 miljard vectoren, geïmplementeerd via een gloednieuwe vierlaagse boomarchitectuur (preview) die de querycomplexite...","url":"https://www.aioga.com/nl/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:10.514Z"},"tr":{"title":"AlloyDB ScaNN Vektör Aramasını 10 Milyar Vektöre Nasıl Genişletiyor","summary":"AlloyDB'nin ScaNN indeksi artık 10 milyar vektörün üzerinde ölçekleri destekliyor; bu, sorgu karmaşıklığını O(N^1/2)'den O(N^1/4)'e düşüren yepyeni dört katmanlı ağaç mimarisi (önizleme) ile uygulanıyor. İç testlerde, bu mimari 10 milyar vektör ölçekte P95 gecikmesini 51 milisaniyeden fazla olmayan ve %95 geri çağırma oranına ulaşır. Bu özellik Hızlı Başlatma Rehberi aracılığıyla kullanılabilir ve yeni kullanıcılar 30 günlük ücretsiz deneme süresinin tadını çıkarabilir. 🔗 Orijinal makaleyi AIHOT üzerinden okuyun · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"AlloyDB ScaNN Vektör Aramasını 10 Milyar Vektöre Nasıl Genişletiyor - Aioga AI Haberleri","description":"AlloyDB'nin ScaNN indeksi artık 10 milyar vektörün üzerinde ölçekleri destekliyor; bu, sorgu karmaşıklığını O(N^1/2)'den O(N^1/4)'e düşüren yepyeni dört katmanlı ağaç mimarisi (öni...","url":"https://www.aioga.com/tr/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:18.312Z"},"vi":{"title":"Cách AlloyDB ScaNN mở rộng tìm kiếm vector lên 10 tỷ vector","summary":"Chỉ số ScaNN của AlloyDB hiện hỗ trợ các quy mô vượt quá 10 tỷ vector, được triển khai thông qua kiến trúc cây bốn lớp hoàn toàn mới (xem trước) giúp giảm độ phức tạp truy vấn từ O(N^1/2) xuống O(N^1/4). Trong kiểm thử nội bộ, kiến trúc này đạt độ trễ P95 không quá 51 mili giây và tỷ lệ thu hồi 95% trên thang đo 10 tỷ vector. Tính năng này có thể được triển khai thông qua Hướng dẫn Bắt đầu Nhanh, và người dùng mới có thể tận hưởng bản dùng thử miễn phí 30 ngày. 🔗 Đọc bài viết gốc qua AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Cách AlloyDB ScaNN mở rộng tìm kiếm vector lên 10 tỷ vector - Tin tức AI Aioga","description":"Chỉ số ScaNN của AlloyDB hiện hỗ trợ các quy mô vượt quá 10 tỷ vector, được triển khai thông qua kiến trúc cây bốn lớp hoàn toàn mới (xem trước) giúp giảm độ phức tạp truy vấn từ O...","url":"https://www.aioga.com/vi/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:19.038Z"},"id":{"title":"Bagaimana AlloyDB ScaNN Memperluas Pencarian Vektor hingga 10 Miliar Vektor","summary":"Indeks ScaNN AlloyDB kini mendukung skala melebihi 10 miliar vektor, yang diimplementasikan melalui arsitektur pohon empat lapis baru (pratinjau) yang mengurangi kompleksitas kueri dari O(N^1/2) menjadi O(N^1/4). Dalam pengujian internal, arsitektur ini mencapai latensi P95 tidak lebih dari 51 milidetik dan laju recall 95% pada skala 10 miliar vektor. Fitur ini dapat diterapkan melalui Panduan Mulai Cepat, dan pengguna baru dapat menikmati uji coba gratis selama 30 hari. 🔗 Baca artikel asli melalui AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Bagaimana AlloyDB ScaNN Memperluas Pencarian Vektor hingga 10 Miliar Vektor - Berita AI Aioga","description":"Indeks ScaNN AlloyDB kini mendukung skala melebihi 10 miliar vektor, yang diimplementasikan melalui arsitektur pohon empat lapis baru (pratinjau) yang mengurangi kompleksitas kueri...","url":"https://www.aioga.com/id/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:26.307Z"},"th":{"title":"AlloyDB ScaNN ขยายการค้นหาเวกเตอร์ไปถึง 10 พันล้านเวกเตอร์ได้อย่างไร","summary":"ดัชนี ScaNN ของ AlloyDB รองรับสเกลที่เกิน 10 พันล้านเวกเตอร์ ซึ่งถูกพัฒนาผ่านสถาปัตยกรรมต้นไม้สี่ชั้นใหม่ (พรีวิว) ที่ช่วยลดความซับซ้อนของคําสั่งจาก O(N^1/2) เหลือ O(N^1/4) ในการทดสอบภายใน สถาปัตยกรรมนี้สามารถทําความหน่วง P95 ได้ไม่เกิน 51 มิลลิวินาที และอัตราการเรียกคืน 95% ที่มาตราส่วน 10 พันล้านเวกเตอร์ ฟีเจอร์นี้สามารถนําไปใช้ผ่านคู่มือเริ่มต้นด่วน และผู้ใช้ใหม่สามารถทดลองใช้งานฟรี 30 วัน 🔗 อ่านบทความต้นฉบับผ่าน AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"AlloyDB ScaNN ขยายการค้นหาเวกเตอร์ไปถึง 10 พันล้านเวกเตอร์ได้อย่างไร - ข่าว AI Aioga","description":"ดัชนี ScaNN ของ AlloyDB รองรับสเกลที่เกิน 10 พันล้านเวกเตอร์ ซึ่งถูกพัฒนาผ่านสถาปัตยกรรมต้นไม้สี่ชั้นใหม่ (พรีวิว) ที่ช่วยลดความซับซ้อนของคําสั่งจาก O(N^1/2) เหลือ O(N^1/4) ในการทด...","url":"https://www.aioga.com/th/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:27.534Z"},"pl":{"title":"Jak AlloyDB ScaNN rozszerza wyszukiwanie wektorowe do 10 miliardów wektorów","summary":"Indeks ScaNN AlloyDB obsługuje obecnie skale przekraczające 10 miliardów wektorów, zaimplementowane za pomocą zupełnie nowej architektury drzewa czterech warstw (podgląd), która zmniejsza złożoność zapytań z O(N^1/2) do O(N^1/4). W testach wewnętrznych ta architektura osiąga opóźnienie P95 nie przekraczające 51 milisekund oraz współczynnik przywołania 95% w skali 10 miliardów wektorów. Tę funkcję można wdrożyć za pomocą Przewodnika Szybkiego Startu, a nowi użytkownicy mogą skorzystać z 30-dniowego bezpłatnego okresu próbnego. 🔗 Przeczytaj oryginalny artykuł za pośrednictwem AIHOT · https://aihot.virxact.com/items/cmt1r7jjh05lgroovc4zwypj6","category":"行业动态","source":"Google Cloud：Databases（RSS","aggregationSource":"Google Cloud：Databases（RSS","pageTitle":"Jak AlloyDB ScaNN rozszerza wyszukiwanie wektorowe do 10 miliardów wektorów - Aioga Wiadomości AI","description":"Indeks ScaNN AlloyDB obsługuje obecnie skale przekraczające 10 miliardów wektorów, zaimplementowane za pomocą zupełnie nowej architektury drzewa czterech warstw (podgląd), która zm...","url":"https://www.aioga.com/pl/news/cmt1r7jjh05lgroovc4zwypj6/","contentTranslated":true,"sourceHash":"fe9ad4397e6fba2a","translatedAt":"2026-08-20T17:02:36.141Z"}},"evidenceTier":"verified-news","reviewStatus":"editorial-selected","indexable":true,"editorialCover":""}}