SAP-C03 · Domain 3: Cost-Optimized Architecture Design · 15% of the exam
Task 3.2: Optimize data transfer and network costs.
What moving data costs: CloudFront and Origin Shield, private connectivity and AZ-aware designs instead of NAT and cross-zone paths, cost-aware replication between Regions, Transit Gateway and PrivateLink pricing, and cross-Region inference for AI workloads.
Study it
Data transfer costs: CloudFront, VPC endpoints, NAT gateways and AZ-aware designs
Network pricing and replication between Regions, including cross-Region inference
Lesson coming
Sample questions
Try each one before opening the answer. Every option is explained, with the AWS documentation page that proves it.
Question 1 · choose 1
In the design of a new data platform, EC2 instances in private subnets will read about 300 TB per month from Amazon S3 and write results to Amazon DynamoDB, both in the same Region. The draft sends this traffic through NAT gateways. The data must not travel over the internet. Which change gives the largest cost reduction?
AReplace the NAT gateways with NAT instances on large EC2 instance types
BAdd gateway VPC endpoints for Amazon S3 and DynamoDB to the route tables of the private subnets
CAdd interface VPC endpoints for Amazon S3 and DynamoDB in each Availability Zone
DMove the instances to public subnets with public IP addresses and reach the services through an internet gateway
Show the answer and why
AReplace the NAT gateways with NAT instances on large EC2 instance types
Incorrect
NAT instances remove the NAT gateway charges but add instance costs and management, and the traffic would still pass through a NAT device.
BAdd gateway VPC endpoints for Amazon S3 and DynamoDB to the route tables of the private subnets
Correct
Gateway endpoints for S3 and DynamoDB carry the traffic privately from the VPC and have no additional charge, so the NAT gateway data processing charges for this traffic disappear.
CAdd interface VPC endpoints for Amazon S3 and DynamoDB in each Availability Zone
Incorrect
Interface endpoints are billed per hour and per GB processed, so they cost more than gateway endpoints for the same traffic.
DMove the instances to public subnets with public IP addresses and reach the services through an internet gateway
Incorrect
This removes the NAT charges but exposes the instances, and the requirement that data does not use the internet path would no longer be met.
For S3 and DynamoDB in the same Region, gateway endpoints are both the private path and the cheapest one.
A video platform serves viewers worldwide through CloudFront from one origin in us-east-1. During popular launches, many edge locations miss at once and the origin is overloaded. Which CloudFront feature reduces load on the origin?
AField-level encryption for viewer requests
BCloudFront Origin Shield
CA shorter cache TTL
DGeographic restrictions by country
Show the answer and why
AField-level encryption for viewer requests
Incorrect
Field-level encryption protects data; it does not reduce origin load.
BCloudFront Origin Shield
Correct
Origin Shield is an additional caching layer that helps minimize the origin's load and improve its availability.
CA shorter cache TTL
Incorrect
Shorter TTLs send more requests to the origin.
DGeographic restrictions by country
Incorrect
Blocking countries reduces audience, not cache misses.
An extra central cache layer in front of the origin is Origin Shield.
A document summarization service runs on EC2 in eu-west-1 and sends large prompts to a model in Amazon Bedrock. Because the model has more capacity in another EU Region, the team configured the SDK to call that Region's Bedrock endpoint directly, and the bill now shows a large inter-Region data transfer line from eu-west-1. The model is offered in eu-west-1 through the EU geographic cross-Region inference profile, and data must stay in the EU. Which change reduces cost while keeping the extra capacity?
ACall the model from eu-west-1 through the EU geographic cross-Region inference profile
BRoute the calls to the other Region's endpoint through a VPC peering connection with that Region
CCompress the prompts in the application before sending them to the other Region's endpoint
DSwitch to the global cross-Region inference profile, which is about 10% cheaper per token
Show the answer and why
ACall the model from eu-west-1 through the EU geographic cross-Region inference profile
Correct
Cross-Region inference has no additional routing cost; the price is based on the Region from which the profile is called, and requests travel on the AWS network within the EU geography.
BRoute the calls to the other Region's endpoint through a VPC peering connection with that Region
Incorrect
Traffic that leaves eu-west-1 for another Region is still billed as inter-Region data transfer out of the source Region, peering or not.
CCompress the prompts in the application before sending them to the other Region's endpoint
Incorrect
Smaller payloads lower the transfer volume somewhat, but every request still pays inter-Region data transfer, while a profile called locally adds no routing cost.
DSwitch to the global cross-Region inference profile, which is about 10% cheaper per token
Incorrect
Global profiles can route requests outside the EU, which breaks the requirement that data stay in the EU.
Cross-Region inference spreads load across Regions without extra routing charges and is priced from the source Region, so calling a geographic profile locally is cheaper than sending traffic to another Region's endpoint yourself.
A microservices platform on Amazon EKS spreads its pods evenly across three Availability Zones for resilience. The services call each other through Kubernetes Services at high volume, and the bill shows large charges for data transfer between Availability Zones because calls often land on pods in another zone. The company wants to keep the three-zone layout and cut this charge without changing application code. Which change meets these requirements?
AMove all pods into one Availability Zone so that every call stays inside that zone
BReplace the Kubernetes Services with a public Application Load Balancer for each service
CTurn on topology aware routing for the Services, with topology spread constraints to balance endpoints per zone
DAdd more replicas of every service in each zone so that a larger share of calls happens to land on a local pod
Show the answer and why
AMove all pods into one Availability Zone so that every call stays inside that zone
Incorrect
This removes the inter-zone charge but gives up the three-zone layout that the company keeps for resilience.
BReplace the Kubernetes Services with a public Application Load Balancer for each service
Incorrect
An internet-facing load balancer adds cost and still spreads requests across zones; it does not keep traffic in the caller's zone.
CTurn on topology aware routing for the Services, with topology spread constraints to balance endpoints per zone
Correct
Topology aware routing gives endpoints zone hints so that kube-proxy prefers endpoints in the caller's zone, and spread constraints keep enough endpoints in each zone for the hints to apply.
DAdd more replicas of every service in each zone so that a larger share of calls happens to land on a local pod
Incorrect
Without topology hints, Services spread calls across endpoints in all zones, so adding replicas evenly does not change the share of calls that cross zones.
Inter-AZ data transfer is billed in both directions. Keeping service calls in the caller's zone with topology aware routing cuts that cost while the workload stays spread across zones.
A company replicates 40 TB of new S3 objects each month from us-east-1 to us-west-2 for disaster recovery, with S3 Replication Time Control and replication metrics turned on. The recovery plan accepts replicas that are a few hours behind, and nobody uses the replication SLA or its metrics. All new objects must still be copied to us-west-2. Which change lowers the replication cost while meeting the recovery plan?
AReplace Cross-Region Replication with S3 Batch Replication jobs that run every hour
BWrite the replicas to the S3 Glacier Deep Archive storage class in us-west-2
CChange the destination storage class to S3 Standard and turn on S3 Transfer Acceleration for the bucket
DTurn off Replication Time Control and its metrics and keep standard Cross-Region Replication
Show the answer and why
AReplace Cross-Region Replication with S3 Batch Replication jobs that run every hour
Incorrect
Batch Replication is meant for existing objects and adds job costs. It is not a cheaper replacement for continuous replication of new objects.
BWrite the replicas to the S3 Glacier Deep Archive storage class in us-west-2
Incorrect
Cheaper replica storage does not remove the RTC charges, and archived replicas would have to be restored before a recovery could use them.
CChange the destination storage class to S3 Standard and turn on S3 Transfer Acceleration for the bucket
Incorrect
Transfer Acceleration speeds up transfers between clients and S3 at an extra charge; it does not reduce replication cost.
DTurn off Replication Time Control and its metrics and keep standard Cross-Region Replication
Correct
Replication Time Control adds its own data transfer charge and replication metrics charges on top of standard replication. Standard Cross-Region Replication still copies every new object.
With S3 Replication Time Control you pay an extra RTC data transfer charge and metrics charges for the 15-minute SLA. When the recovery plan does not need that SLA, standard replication does the job for less.