Opensearch Domain - Sizing recommendations

Customers that use Amazon Connect will be hosting the AWS-Managed Opensearch domain on their AWS.

This document would show the sizing recommendations based on number of active agents on their Amazon Connect instance.

Caveat

All Opensearch Domains here will be running the Elasticsearch 7.9 engine.

Please note as well, that the number of active agents is not the only factor for Opensearch domain CPU/Memory load.
Dashboard complexity, peak concurrent dashboard users can also cause considerable load on the domain.
Which is why we recommend customers to also monitor the following metrics on their Opensearch domain:

• Data Nodes CPU Utilization and JVM Memory Pressure

  • Open

    

     

• Write / Search Thread Pools

  • Open

    

     

If the CPU/Memory Pressure is reaching above 90%, or if the Write/Search thread pools are showing queues more than 300 over prolonged periods of time (especially during peak hours), we recommend increasing the Data Node size, even if you don’t have as many active agents as in the recommendations below.

Less than 1,000 Agents (Standard Deployment)

This the default sizing typically used for new customers.

• Standard

  • Master Nodes: None

  • Data Nodes

    • 1 Data Node

    • Size: m6g.large

    • Storage: 100GB

• If customer requires HA/Multi-AZ for their deployment

  • This configuration would be able to handle loss of an AZ

  • Master Nodes

    • 2 Master Nodes

    • Size: c6g.large

  • Data Nodes

    • 2 Data Nodes

    • Size: m6g.large

    • Storage: 100GB

1,000 - 3,000 Agents

• Standard

  • Master Nodes: None

  • Data Nodes

    • 1 Data Node

    • Size: m6g.xlarge

    • Storage: 200GB

• If customer requires HA/Multi-AZ for their deployment

  • This configuration would be able to handle loss of an AZ

  • Master Nodes

    • 2 Master Nodes

    • Size: c6g.large

  • Data Nodes

    • 2 Data Nodes

    • Size: m6g.xlarge

    • Storage: 200GB

3,000 - 8,000 Agents

• Due to the increasingly large volume of data, we recommend a minimum HA configuration

• Master Nodes

  • 2 Master Nodes

  • Size: m6g.large

• Data Nodes

  • 2 Data Nodes

  • Size: m6g.2xlarge

  • Storage: 300GB

Greater than 8,000 Agents

• Due to the increasingly large volume of data and write/read activity, we recommend a 3-node configuration

• Master Nodes

  • 3 Master Nodes

  • Size: m6g.large

• Data Nodes

  • 3 Data Nodes

  • Size: m6g.4xlarge

  • Storage: 500GB