Configure your Splunk Observability Cloud account to collect Amazon Bedrock metrics

Learn how to configure your Splunk Observability Cloud account to collect Amazon Bedrock metrics.

You can monitor the performance of Amazon Bedrock applications by configuring your Amazon Bedrock applications to send metrics to Splunk Observability Cloud. This solution creates a cloud connection in your Splunk Observability Cloud account that collects metrics from AWS CloudWatch.

Complete the following steps to collect metrics from Amazon Bedrock applications.

  1. Connect AWS to Splunk Observability Cloud. For more information on the connection methods and instructions for each method, see Available options to connect with AWS.
    To gain permission to collect all Amazon Bedrock attributes, add the following permissions to your AWS IAM policy JSON snippet:
    JSON 
    "bedrock:ListTagsForResource",
"bedrock:ListFoundationModels",
"bedrock:GetFoundationModel",
"bedrock:ListInferenceProfiles"
  2. To monitor Amazon Bedrock metrics with Splunk Observability Cloud, run your applications that use Amazon Bedrock models.

Metrics

The following metrics are available for Amazon Bedrock applications. For more information on these metrics, see Monitoring the performance of Amazon Bedrock in the AWS documentation and View organization metrics for Splunk Observability Cloud.

These metrics are considered custom metrics in Splunk Observability Cloud.

Metric name Type Unit Description
InputTokenCount counter count Input tokens consumed.
OutputTokenCount counter count Output tokens consumed.
Invocations counter SampleCount The number of successful requests made to Amazon Bedrock API operations.
InvocationLatency histogram ms The time taken for an invocation, specifically the latency of the requests to the Amazon Bedrock models, typically in milliseconds.
sf.org.num.awsServiceAuthErrorCount gauge count

Total number of authentication errors thrown by AWS services.

  • Dimension(s): integrationId, orgId, namespace (AWS Cloudwatch namespace), clientInterface (AWS SDK Interface), method (AWS SDK method)

  • Data resolution: 1 second
sf.org.num.awsServiceCallThrottles gauge count

Number of calls made to the Amazon API that are being throttled by AWS because you have exceeded your AWS API call limits.

  • Dimension(s): orgId, namespace (AWS service, such as AWS/Cloudwatch), method (the API being called, such as getMetricStatistics)

  • Data resolution: 5 seconds
sf.org.num.awsServiceCallThrottledRetryFailures gauge count Number of AWS service call throttle retry failures.
sf.org.num.awsServiceCallCountExceptions gauge count

Number of calls made to the Amazon API that threw exceptions.

  • Dimension(s): orgId, namespace (AWS service, such as AWS/Cloudwatch), method (the API being called, such as getMetricStatistics), location (where the exception occurred, either client or server)

  • Data resolution: 5 seconds
sf.org.num.awsServiceCallCount gauge count

Total number of calls made to the Amazon API.

  • Dimension(s): orgId, namespace (AWS service, such as AWS/Cloudwatch), method (the API being called, such as getMetricStatistics)

  • Data resolution: 5 seconds

Attributes

The following resource attribute is available for Amazon Bedrock applications.

OpenTelemetry name Display name Description
id id The AWS unique identifier.

Next steps

How to monitor your AI components after you set up Observability for AI.

After you set up data collection from supported AI infrastructure components to Splunk Observability Cloud, the data populates built-in experiences that you can use to monitor and troubleshoot them.

The following table describes the tools you can use to monitor and troubleshoot your AI infrastructure components.
Do this With this tool Link to documentation
Orient and explore different layers of your AI tech stack. Built-in navigators
Assess service, endpoint, and system health at a glance. Built-in dashboards