Domain 2 of 4 · Chapter 5 of 6

Performance for new solutions

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing AWS Certified Solutions Architect - Professional premium course — no separate purchase.

Included in this chapter:

  • Compute selection: families, Graviton, burstable, serverless
  • Storage performance: EBS, instance store, EFS, FSx, S3
  • Database performance: engines, replicas, RDS Proxy
  • Caching, edge, networking, and decoupling for scale
  • Exam patterns: reading a performance question right

Caching layers for new solutions: which cache where (SAP-C02)

DimensionCloudFrontElastiCache (Redis/Memcached)DynamoDB Accelerator (DAX)API Gateway caching
What it cachesHTTP responses / static + cacheable dynamic content at the edgeArbitrary app data: query results, sessions, leaderboardsDynamoDB GetItem/Query/Scan resultsAPI responses keyed by request parameters
Where it sitsAWS edge locations (global, closest to user)In-VPC, in-Region, near computeIn-VPC, in front of one DynamoDB tableAt the API Gateway stage
Latency it targetsInternet RTT to originDatabase/query round-trips (sub-ms in-memory)DynamoDB read latency (ms to microseconds)Repeat backend invocations
Coupling to data storeOrigin-agnostic (S3, ALB, any HTTP origin)Application manages reads/writes + invalidationDynamoDB-only, write-through, transparent SDKBackend-agnostic per method
Typical fitGlobal content delivery, media, websitesRelational read offload, sessions, pub/subHot-key DynamoDB read amplificationIdempotent GETs with reusable responses

Decision tree

Where is the latency bottleneck? Compute-bound Tightly-coupled HPC / distributed training? Data-tier Read load or connection churn? Global users Content cacheable over HTTP? Yes Cluster placement group + EFA (OS-bypass, SRD) No Graviton + right family; Lambda / Fargate / Batch Connections RDS Proxy (pool connections) Reads Read replicas + ElastiCache; DAX for DynamoDB hot keys Yes CloudFront (edge CDN cache) No (TCP/UDP) Global Accelerator (anycast, backbone) Spiky load? Decouple first — buffer with SQS / Kinesis, scale consumers on backlog before scaling the synchronous tier.
Cheat sheet

Cheat sheet

  • Choose the instance family by the binding resource
  • Burstable T-family throttles to baseline when credits run out
  • Prefer Graviton for ~20% lower cost (up to 60% less energy)
  • Lambda hard limits: 15 min, 10 GB, 6 MB / 1 MB
  • Cluster placement group + EFA for tightly-coupled HPC/ML
  • gp3 provisions IOPS/throughput independently of size
  • io2 Block Express ceiling: 256k IOPS, 4,000 MiB/s
  • EFS Elastic throughput auto-scales for spiky workloads
  • FSx for Lustre is the HPC/ML parallel file system
  • S3 scales per prefix: 3,500 writes / 5,500 reads
  • Pick the database engine by access pattern
  • DAX caches DynamoDB reads to microseconds
  • Read replicas scale reads; Aurora lag is sub-100 ms
  • RDS Proxy solves connection churn, not read volume
  • Aurora Serverless v2 scales fine-grained; v1 is retired
  • ElastiCache: Redis for features, Memcached for simple cache
  • CloudFront vs Global Accelerator: cacheable HTTP vs TCP/UDP
  • Placement groups: cluster vs spread vs partition
  • Decouple with SQS/Kinesis to absorb spikes
  • Availability features don't add read throughput
  • Place CloudFront Origin Shield in the Region closest to the origin
  • Front multiple CDNs with CloudFront + Origin Shield to consolidate origin fetches
  • ElastiCache Redis cluster mode scales horizontally by sharding
  • Redis hash tags pin related keys to one slot for multi-key operations

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. Burstable performance instances
  2. Placement groups
  3. Elastic Fabric Adapter
  4. AWS Graviton Processor
  5. Lambda quotas
  6. AWS Fargate for Amazon ECS
  7. What is AWS Batch?
  8. General Purpose SSD (gp2 and gp3) volumes
  9. Provisioned IOPS SSD (io1 and io2) volumes
  10. Amazon EFS performance
  11. What is Amazon FSx for Lustre?
  12. Amazon EC2 instance store
  13. Best practices design patterns: optimizing Amazon S3 performance
  14. What is Amazon DynamoDB?
  15. What is Amazon ElastiCache?
  16. Replication with Amazon Aurora
  17. Using Amazon RDS Proxy
  18. Performance and scaling for Aurora Serverless v2
  19. In-memory acceleration with DynamoDB Accelerator (DAX)
  20. What is Amazon CloudFront?
  21. What is AWS Global Accelerator?
  22. What is Amazon Simple Queue Service?
  23. What is Amazon SNS?
  24. What is Amazon Kinesis Data Streams?
  25. Attach a volume to multiple instances with Amazon EBS Multi-Attach
  26. Uploading and copying objects using multipart upload
  27. Configuring fast, secure file transfers using Amazon S3 Transfer Acceleration
  28. What is Amazon Aurora?
  29. Amazon DynamoDB on-demand capacity mode pricing
  30. Working with DB instance read replicas
  31. Caching API responses with API Gateway
  32. Enhanced networking with ENA
  33. Network maximum transmission unit (MTU) for your EC2 instance
  34. Performance Efficiency Pillar - AWS Well-Architected Framework Whitepaper