Cost Optimization Pillar
The ability to run systems to deliver business value at the lowest price point.
Official Questions
Best Practices
High Risk if Missing
AWS Verbatim
Data transfer charge traps (cross-AZ $0.02/GB, NAT Gateway $0.045/GB vs free Gateway Endpoints), right-sizing, storage lifecycle policies, and serverless idle efficiency.
In AWS Well-Architected, questions do not have single-choice trick answers. Every listed Best Practice represents an official architectural answer you must incorporate into your workload. When asked these questions in an Amazon System Design interview, your score is evaluated by how many of these best practices you proactively articulate and defend with trade-offs.
Verbatim Questions & Architecture Answers
How do you implement cloud financial management?
Establish cross-functional cost ownership, define cloud budgets, and track unit metrics.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Assign cloud budget ownership directly to engineering teams.
Track unit cost (e.g. cost per active user or cost per 1,000 orders).
Ensures architectures are measured on cost-efficiency rather than raw over-provisioning.
Fosters a frugality culture aligned with Amazon's Leadership Principles.
Clear tagging enables automated cost allocation across microservice owners.
How do you govern usage?
Use Service Control Policies (SCPs) and resource tagging to enforce cost guardrails.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Enforce SCPs that prevent provisioning of expensive unapproved instance families.
Require mandatory tags on all infrastructure resources to trace spending.
Guardrails keep production and development environments cleanly segmented.
Stops accidental rogue provisioning of expensive multi-thousand-dollar GPU fleets.
Tagging enables automated cost reporting and resource lifecycle tracking.
How do you monitor your cost and usage?
Deploy AWS Budgets, Cost Anomaly Detection, and detailed billing reports.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Analyze AWS Cost & Usage Reports (CUR) in Amazon Athena.
Receive immediate Slack/SNS alerts when daily spending deviates from expected baselines.
Real-time anomaly detection alerts engineers to infinite loops or DDoS traffic surges.
Prevents surprise month-end AWS billing shock.
Automated anomaly alerts reduce manual financial auditing toil.
How do you decommission resources?
Identify and purge orphaned EBS volumes, unattached Elastic IPs, and stale data via TTL.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Automate discovery and deletion of unattached volumes and abandoned test environments.
Enable DynamoDB TTL and S3 Lifecycle rules to prune expired records automatically.
Pruning stale items keeps database indexes compact, maintaining single-digit millisecond lookups.
Eliminates paying for gigabytes of dead storage and abandoned resources.
Zero manual intervention: DynamoDB TTL deletes expired rows in the background at zero write cost.
How do you evaluate cost when you select services?
Analyze Total Cost of Ownership (TCO), comparing managed services vs self-hosted infrastructure.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Account for engineer maintenance hours, patching, and backup operational costs.
Select DynamoDB/SQS over self-hosted Cassandra/Kafka when operational overhead exceeds compute savings.
Managed cloud services provide built-in multi-AZ replication and automated scaling.
Serverless pricing is higher per unit at massive scale, but saves hundreds of thousands in DevOps salaries.
Eliminates undifferentiated heavy lifting: no OS patching, no disk rebalancing, no hardware maintenance.
How do you meet cost targets when you select resource type, size and number?
Right-size compute, memory, and database instances based on observed utilization metrics.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Downsize instances running below 20% average CPU/memory utilization.
Match instance counts directly to workload demand curves.
Right-sizing maintains adequate headroom for traffic bursts while eliminating excess slack.
Reduces compute spend by 30-50% across development and production fleets.
AWS Compute Optimizer provides automated ML recommendations for instance resizing.
How do you use pricing models to reduce cost?
Blend On-Demand, Compute Savings Plans, Reserved Instances, and Spot Instances.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Commit to 1 or 3-year Compute Savings Plans for up to 66% discount on baseline load.
Run fault-tolerant stateless workers and queue processors on Spot for up to 90% savings.
Spot Instances can be interrupted with 2-minute notice; workers must be stateless and idempotent.
Blended pricing models deliver the lowest possible cloud compute spend at scale.
EC2 Auto Scaling Mixed Instances Groups automatically manages the Spot-to-On-Demand ratio.
How do you plan for data transfer charges?
Avoid hidden AWS data transfer traps by minimizing cross-AZ traffic, using VPC Gateway Endpoints, and deploying CloudFront.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Avoid cross-AZ traffic ($0.02/GB round-trip) by maintaining AZ-affinity between app servers and caches.
Route S3 and DynamoDB calls over free internal AWS endpoints to eliminate NAT Gateway fees ($0.045/GB).
CloudFront edge caching offloads origin egress and provides discounted data-out tiers.
Compress payloads (Gzip, Protobuf) before sending over cross-region links ($0.02/GB).
VPC Gateway Endpoints provide lower latency by bypassing public NAT Gateway proxy hops.
Can save tens of thousands of dollars per month on high-throughput platforms.
Simple one-time VPC route table configuration with zero application code changes.
How do you manage demand, and supply resources?
Use asynchronous buffer queues to smooth traffic peaks and throttle non-essential requests.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Use Amazon SQS to buffer spikes and process work with a right-sized steady worker pool.
Ensure auto-scaling cooldown timers and termination policies quickly scale down unused capacity.
Queuing converts synchronous load shedding into delayed processing, preserving request completion.
Enables sizing worker pools for average throughput rather than peak burst capacity.
Decouples request intake from background processing pipelines.
How do you evaluate new services?
Continuously benchmark new AWS service tiers and instance families to harvest efficiency improvements.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Migrate legacy x86 workloads to Graviton3/Graviton4 to achieve immediate 20% cost reductions.
Newer instance types typically provide higher IPC and faster memory bandwidth.
Direct unit price discounts for upgraded architectural generations.
Requires validating multi-architecture Docker container builds (arm64).
How do you evaluate the cost of effort?
Stop spending money on undifferentiated heavy lifting; evaluate the true engineering cost of maintaining custom solutions.
Recommended Answers (Official AWS Best Practices Checklist)
Every best practice below is an official answer to incorporate into your system design:
Avoid custom bespoke infrastructure when managed cloud primitives provide 95% of capability.
Cloud-native managed primitives benefit from continuous performance upgrades maintained by AWS.
Saves massive engineering team hours from maintenance toil.
Keeps the codebase focused on core business logic rather than distributed system plumbing.