Skip to main content

Machine learning

Advanced Managed Search (OpenSearch) provides various machine learning–based features such as data analysis, anomaly detection, vector search, and semantic search through the OpenSearch Machine Learning (ML) capability.
This page briefly explains the overview of machine learning features provided in Advanced Managed Search and considerations when operating them.

Overview of machine learning features​

The machine learning capability of OpenSearch allows models to run within the cluster and supports combining machine learning–based processing with search and analytics workflows.
Advanced Managed Search supports all machine learning capabilities provided by OpenSearch, enabling the following use cases.

Tasks you can perform with machine learning​

By using the machine learning capabilities of Advanced Managed Search, you can perform tasks such as the following.

  • Anomaly detection

    • Detect anomaly patterns based on time-series data
    • Identify abnormal behavior in logs, metrics, and traffic data

  • Natural language processing (NLP)–based analysis

    • Generate text embeddings
    • Perform semantic search and query expansion

  • Integration with vector search

    • Generate embeddings using machine learning models
    • Perform similarity search combined with vector search capabilities

  • Model management and execution

    • Register and manage machine learning models
    • Invoke models within search and analytics pipelines

Machine learning workflow​

Machine learning capabilities are typically used in the following workflow.

  1. Prepare model

    • Prepare a pre-trained model or a model created externally

  2. Register model

    • Register the model using the OpenSearch machine learning capability

  3. Run model

    • Apply the model to search, analytics, or anomaly detection tasks

  4. Use results

    • Improve search accuracy
    • Generate alerts and analysis based on anomaly detection results

Detailed usage of machine learning features​

For detailed usage methods and examples such as machine learning APIs, model registration, anomaly detection configuration, and vector embedding generation, refer to the OpenSearch official documentation.

Notes when using machine learning​

In Advanced Managed Search, the following considerations must be observed to ensure stable operation of machine learning features.

Caution

Do not use the cluster setting plugins.ml_commons.only_run_on_ml_node: false in a production environment.

This option may reduce the stability of data nodes. Instead, use it through connector configuration.

note

Using Connector is recommended in production environments.

When directly invoking models or integrating external systems in production environments, using a Connector is recommended to separate machine learning workloads and enhance cluster stability and security.

For machine learning tutorials, see AI with OpenSearch.