Reducing cross-region data transfer costs

WHAT THIS TESTS This checks whether you can diagnose data transfer billing, which is one of the least intuitive cloud cost drivers, and apply targeted architectural fixes.

A GOOD ANSWER COVERS First, enumerate the chargeable paths. Cross-region traffic between us-east-1 and us-west-2 is billed per gigabyte in both directions and is usually the biggest culprit when components span regions. Cross-AZ traffic within a region is also metered, so a chatty service and database in different AZs add up. Egress to the public internet, often routed through a NAT gateway, carries both NAT processing charges and egress fees. Calls to managed services over their public endpoints can leave the VPC unnecessarily. To reduce cost, co-locate tightly coupled, high-volume components in the same region and ideally the same AZ. Add Gateway VPC endpoints for S3 and DynamoDB and Interface endpoints for other services so that traffic stays on the private network and bypasses NAT and internet egress. Front user-facing content with a CDN like CloudFront to cache at the edge and cut origin egress. Compress and batch payloads to move fewer bytes.

COMMON WRONG ANSWERS Believing inbound data transfer is the problem when ingress is generally free. Ignoring NAT gateway egress, a frequent silent cost. Assuming intra-AZ traffic is the driver. Proposing only a CDN while leaving the expensive cross-region backend chatter untouched.

LIKELY FOLLOW-UPS How do you measure transfer by source. When is replicating data across regions cheaper than transferring it repeatedly. How do VPC peering and Transit Gateway costs compare.

ONE CONCRETE EXAMPLE An app server in us-east-1 was querying a database in us-west-2 on every request. Moving a read replica into us-east-1 eliminated the cross-region per-request transfer, and adding an S3 Gateway endpoint removed NAT egress for object reads, sharply lowering the bill.