AMI metering for usage for complex deployment model

Question

Hello, we are preparing to offer an AMI based product where we track usage based upon end device count information.  Generally understand the process to update this via the API but wanted to see if anyone had some guidance as to what we need to do to make sure we account correctly.
We support a level of deployment redundancy, where EC2 instances in separate regions can be used by the end device to connect/communicate, and the instances share a common separate database.  It's the total provisioned device count in the database that we need to bill on, not the specific number of devices that may be connected one or the other EC2 nodes.  If we have each of the EC2 instances report the provisioned total from the database each hour, there is the possibility (in our mind) that there will be two duplicated usage records, and hence flow on to incorrect billing of the customer.  Indeed, there could be up to 6 regional EC2 instances all acting as part of a global, single cluster.
Any thoughts on how we need to set up the AMI/metering would be very much appreciated.

re:Post Agent · Accepted Answer

To address your complex deployment model for AMI metering, you'll need to implement a careful strategy to avoid duplicate usage records and ensure accurate billing. Here are some recommendations:

1. Designate a Primary Reporter: Choose one EC2 instance as the primary reporter for usage. This could be based on a leader election process or a predetermined order (e.g., alphabetical by region). Only this instance should report the total provisioned device count from the shared database.

2. Implement a Coordination Mechanism: Use your shared database or a separate coordination service (like Amazon DynamoDB) to track which instance has reported usage for each hour. This prevents multiple instances from reporting the same data.

3. Use Unique Identifiers: When calling the MeterUsage API, include a unique identifier for each metering record. This helps ensure idempotency and prevents duplicate records even if called multiple times.

4. Handle Failover: Implement a failover mechanism so that if the primary reporting instance fails, another instance can take over the reporting responsibility.

5. Consistent Timestamp Usage: Ensure all instances use the same time source (like Network Time Protocol) to maintain consistency in reporting periods across regions.

6. Error Handling and Retries: Implement robust error handling and retry logic. If a metering call fails, log the error and attempt to retry at the next reporting interval.

7. Audit Trail: Maintain a detailed audit trail of all metering calls in your shared database. This will help you reconcile reported usage with actual usage if discrepancies arise.

8. Regular Reconciliation: Periodically reconcile your internal usage records with the billing information from AWS Marketplace to ensure accuracy.

Remember, you can only report usage once per hour for each dimension, and you should report usage for the previous hour. The MeterUsage API is idempotent for identical requests, which helps prevent duplicate records if you need to retry a call.

By implementing these strategies, you should be able to accurately meter and bill for your multi-region, database-backed AMI product based on the total provisioned device count, while avoiding duplicate reporting issues.  
**Sources**  
[MeterUsage - AWS Marketplace](https://docs.aws.amazon.com/marketplace/latest/APIReference/API_marketplace-metering_MeterUsage.html)  
[Configuring custom metering for AMI products with AWS Marketplace Metering Service - AWS Marketplace](https://docs.aws.amazon.com/marketplace/latest/userguide/custom-metering-with-mp-metering-service.html)

AMI metering for usage for complex deployment model

Relevant content