2.5 Performance Objectives

Compute Performance

Instance Families (Cheatsheet)

FamilyTypeUse Case
CCompute OptimizedBatch processing, Media transcoding, HPC.
R / XMemory OptimizedIn-memory DBs (Redis), Spark, Real-time Big Data.
I / DStorage OptimizedNoSQL (Cassandra), Data Warehousing.
P / GAcceleratedML Training, Graphics rendering.
MGeneral PurposeApp servers, microservices.

Placement Groups

  • Cluster: Pack instances close together (Single AZ). Low latency.
  • Spread: Place instances on distinct hardware racks. Max 7 per AZ. High durability.
  • Partition: Spread instances across logical partitions. Used for Hadoop/HDFS/Kafka.

Caching Strategy

Amazon ElastiCache

  • Redis: Advanced data structures, Pub/Sub, Persistence, HA (Multi-AZ), Clustering.
  • Memcached: Simple Key-Value, Multithreaded, Sharding only (No Replication/HA).

Amazon CloudFront

  • Function: Content Delivery Network (CDN). Caches static/dynamic content at Edge.
  • Lambda@Edge: Run code at the edge to customize content (e.g., image resizing, header manipulation).
  • Origin Shield: Extra caching layer between Edge and Origin to reduce load.

WARNING

Exam Gotcha: Use Global Accelerator for non-HTTP (TCP/UDP) traffic or when you need static IPs. Use CloudFront for HTTP/S content caching and delivery.

Database Performance

Read Replicas

  • Function: Offload Read content. Async.
  • Engines: MySQL, PostgreSQL, MariaDB, Aurora, Oracle (some versions).

DynamoDB Accelerator (DAX)

  • Function: In-memory cache for DynamoDB. Microsecond latency. No code changes required (Write-Through/Read-Through).