DynamoDB Capacity Mode Cost Thresholds for Low-Traffic Workloads

Optimizing costs for low-traffic DynamoDB tables requires a choice between On-Demand and Provisioned capacity modes. While On-Demand mode removes the need for throughput management by charging per request, Provisioned mode allows for a fixed cost ceiling through Read Capacity Units (RCUs) and Write Capacity Units (WCUs), potentially supplemented by Auto Scaling.

The decision becomes complex when considering the Standard-Infrequent Access (Standard-IA) table class, which lowers storage costs but increases the cost per request. For workloads with minimal but unpredictable traffic, the trade-off between the request-based pricing of On-Demand and the baseline costs of Provisioned mode is not always linear.

• At what specific request volume does Provisioned capacity with Auto Scaling become more cost-effective than On-Demand mode for low-traffic tables?
• How does the higher request cost of the Standard-IA table class shift the break-even point between these two capacity modes?