AWS 启动 S3 矢量器,这是一种无服务器服务,储存多达20万亿矢量,并快速、低成本的AI查询支持。 AWS launches S3 Vectors, a serverless service storing up to 20 trillion vectors with fast, low-cost AI query support.

亚马逊网络服务已推出亚马逊S3矢量器,一般可用,使客户能够储存和查询每个指数高达20亿矢量的矢量——比预览阶段多40倍——并在一个桶里储存和查询多达20万万亿矢量,从而消除了碎片化的需要。 Amazon Web Services has launched Amazon S3 Vectors in general availability, enabling customers to store and query up to 2 billion vectors per index—40 times more than the preview phase—and up to 20 trillion vectors in a single bucket, eliminating the need for sharding. 查询性能现在提供不到一秒钟,用于不经常的搜索,100毫秒以下用于高频查询,支持实时人工智能应用。 Query performance now delivers under one second for infrequent searches and under 100 milliseconds for high-frequency queries, supporting real-time AI applications. 每份查询的结果增加到100份,每秒写1 000份PUT交易。 Results per query have increased to 100, and write throughput reaches 1,000 PUT transactions per second. 完全无服务器服务与亚马逊贝德洛克知识库和开放搜索整合,支持AI工作流程,如RAG和代理开发,在14个AWS地区提供。 The fully serverless service integrates with Amazon Bedrock Knowledge Base and OpenSearch, supports AI workflows like RAG and agent development, and is available in 14 AWS Regions. 与专门的病媒数据库相比,它提供了高达90%的成本节约。 It offers up to 90% cost savings compared to dedicated vector databases.