Polling (default)

1. Your metrics are not time-sensitive, and delays are acceptable

2. You need to use tag filtering or poll rate to manage cost

• Easiest method to set up

• Highest granularity to manage cost through tag filtering or by prolonging polling intervals

• Applies to selected AWS regions under the AWS account

• Metrics might be delayed since they’re are polled at intervals

• If polling intervals are short, this method might be more expensive

Streaming (Splunk-managed)

1. You’re looking for real-time metrics

2. You need to sync metrics from multiple AWS regions

• Access metrics faster and at scale through Kinese Data Firehose streams

• Less expensive than polling when you need close to real-time metrics

• Manage stream filters in Splunk

• Applies to all AWS regions under the AWS account

• Harder to manage costs due to lack of tag filtering and control of streaming input

• You need to deploy CloudFormation templates

Streaming (AWS-managed)

You’re looking for real-time metrics

• Access metrics faster and at scale through Kinese Data Firehose streams

• Less expensive than polling when you need close to real-time metrics

• Direct control over the metric stream set-up

• Harder to manage costs due to lack of tag filtering and control of streaming input

• You need create metric streams through the CloudWatch console for each AWS region