4.3 Eliminating NAT Transfer Fees with Gateway Endpoints & Serverless Databases

🎯 Key AWS SAA-C03 Architectural Takeaways

• Amazon S3 and DynamoDB VPC Gateway Endpoints are completely free of charge and eliminate NAT Gateway hourly and data processing fees ($0.045/GB).
• Amazon Aurora Serverless v2 scales compute in fine-grained Aurora Capacity Units (ACUs) in fractions of a second, saving up to 90% during idle intervals.
• Amazon DynamoDB On-Demand mode charges strictly per read/write request, making it the most cost-effective choice for tables with low, sporadic, or unpredictable traffic.
• Inter-AZ data transfer within a VPC incurs $0.01/GB in each direction; keeping high-volume communication within the same AZ saves network egress costs.

Network architecture decisions have significant cost ramifications. NAT Gateways deployed in public subnets charge an hourly availability rate as well as a data processing fee ($0.045 per gigabyte) for all traffic traversing them. When private analytics fleets download terabytes of data from Amazon S3, NAT Gateway processing charges quickly escalate into thousands of dollars per month.

Creating an Amazon S3 Gateway VPC Endpoint completely eliminates these data transfer costs. Gateway Endpoints modify VPC route tables to direct S3 and DynamoDB traffic directly across AWS internal networking. S3 Gateway Endpoints are completely free of charge, require no NAT Gateway bandwidth, and provide higher throughput and lower latency.

In the database tier, traditional provisioned instances must be sized for peak theoretical load, wasting compute budget during off-peak hours. Amazon Aurora Serverless v2 automatically scales database compute capacity up and down in fine-grained increments (Aurora Capacity Units, or ACUs) in fractions of a second. When traffic drops overnight or on weekends, capacity scales down to minimal baseline, cutting idle database spend dramatically.

Similarly, Amazon DynamoDB On-Demand capacity mode offers true pay-per-request pricing for NoSQL workloads with unknown or bursty access patterns. Instead of paying 24/7 for fixed provisioned Read and Write Capacity Units (RCUs and WCUs), On-Demand mode charges solely for actual requests executed, scaling instantly up to thousands of requests per second and dropping to zero cost during idle periods.

⚠️ Common Pearson VUE / AWS Exam Traps

• Routing high-volume S3 or DynamoDB traffic from private subnets through a NAT Gateway instead of using a free VPC Gateway Endpoint.
• Over-provisioning fixed DynamoDB RCU/WCU on applications that sit idle 95% of the time instead of using On-Demand capacity mode.
• Provisioning peak-capacity RDS database instances for internal business applications used for only 2 hours per day.