A company runs tests on an application that uses an Amazon DynamoDB table for 4 hours once a week. The company knows the number of read and write operations per second during the tests and does not use DynamoDB otherwise. A solutions architect needs to optimize table cost. Which solution meets these requirements?

The correct answer is to choose provisioned mode and update the read and write capacity units appropriately. DynamoDB provisioned mode allows you to specify the exact read and write capacity units (RCUs and WCUs) needed, which is ideal when you have predictable usage patterns, like the weekly 4-hour test. You can scale up for the test and scale down afterward to minimize costs. On-demand mode is more expensive for predictable workloads because it charges per request, which would be higher than provisioning for a known, consistent spike. Purchasing reserved capacity for 1-year or 3-year terms is not suitable here because reserved capacity is designed for continuous, predictable high usage over long periods, not for a brief, weekly spike. The table is not used otherwise, so reserving capacity would lead to significant unused capacity and unnecessary costs.