【全新发布】数小时内构建 GPU 加速、自动优化、十亿级规模的 VectorDB.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Dylan TongProduct Lead,AI and vectors Amazon OpenSearch ServiceAWSBuild GPU-boosted,auto-optimized

2、 billion-scale VectorDBs in hoursVamshi Nakkirtha Senior Manager,Software Dev,Vector SearchAmazon OpenSearch ServiceAWSSession:ANT213 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AWS servicesManaged ClustersAmazon OpenSearch Service ServerlessOpen SourceOpenSearchSearch and Ana

3、lyticsEngineOpenSearch:AWS and Open SourceIngestAmazon OpenSearch IngestionVisualizeOpenSearch UIUse CasesLog Analytics,ObservabilitySearch:Keyword,Vector,Hybrid 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Why Vector Search?1Improves search quality(relevance)2Versatile support

4、 forcontent types3DiverseUse Cases4Agentic(AI)AppEnabler 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Vector Search JourneyFirst ReleaseFeb.20202024Cost Optimizations:Automated QuantizationTiered Vector StorageDisk-optimized EngineAI Native:AI enrichments2023Serverless Amazon B

5、edrock KBHybrid SearchAI Native:Vector Generation2021OpenSearch 1.0:(Elastic 7.10 and Apache Lucene Fork)FAISS 2025Cost Optimizations:S3 Vectors IntegrationIntel AVX-512 Graviton NeonStorage ReductionAI Native:Search AI and Agentic FlowsAutomatic Semantic EnrichmentMCP Server IntegrationEmpower Scal

6、e:Managed Vector IngestionGPU-accelerated Vector IndexingVector Auto-optimize 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Customers trending to