Digital Garden

Amazon OpenSearch Service

Properties1
tagsaws/analytics, cloudpractitioner, aipractitioner

2 min read

Amazon OpenSearch Service

Amazon OpenSearch Service is a managed search and analytics engine from Amazon Web Services used to search, analyze, and visualize textual, numerical, geospatial, structured, and unstructured data. It runs OpenSearch, an open-source search-and-analytics engine forked from Elasticsearch in 2021, and was itself renamed from Amazon Elasticsearch Service (Amazon ES) on September 8, 2021; the service still supports running legacy Elasticsearch OSS versions on existing domains.

AI Practitioner focus

  • OpenSearch supports full-text, semantic, hybrid, and vector similarity search, making it a common vector store/retrieval engine for RAG.
  • A RAG system embeds chunks and queries, retrieves/reranks relevant results, and sends evidence to a foundation model; OpenSearch retrieves evidence but does not generate the final answer.

Key points

  • Managed engine covering textual, numerical, geospatial, structured, and unstructured data; runs the open-source OpenSearch engine (forked from Elasticsearch 7.10.2) as well as legacy Elasticsearch OSS versions.
  • Renamed from Amazon Elasticsearch Service (Amazon ES) to Amazon OpenSearch Service on September 8, 2021; API operations, instance type names, and CloudWatch metrics were renamed to drop “Elasticsearch” in favor of engine-agnostic names.
  • Paired with OpenSearch Dashboards — the renamed successor to Kibana — for visualization, and Logstash for ingestion. After upgrading a domain from Elasticsearch to OpenSearch, the /_plugin/kibana endpoint becomes /_dashboards.
  • Launches cluster resources, detects failed nodes, and automatically replaces them.
  • Cluster can be scaled with a few clicks in the console.
  • Integrates with Amazon CloudWatch, Amazon CloudTrail, Amazon Kinesis Data Streams, Amazon S3, AWS Identity and Access Management (IAM), AWS Lambda, and Amazon DynamoDB.
  • Amazon OpenSearch Serverless is an on-demand, auto-scaling deployment option that decouples indexing from search, removes cluster capacity management entirely, and suits infrequent, intermittent, or unpredictable workloads.

Pricing

  • Charged for each hour of use of the EC2 instances and storage volumes attached to the cluster.
  • No charge for data transfer between Availability Zones.

Sources

Graph View

Backlinks