AWS SOLUTIONS ARCHITECT ASSOCIATE (SAA-C03)
Chapter 5: Deep-Dive into Compute
EC2, Containers, Serverless & Pricing Strategies
Amazon EC2 Instance Families
General Purpose (T, M)
Balanced CPU/RAM. Best for web servers, small DBs, and dev environments.
Balanced CPU/RAM. Best for web servers, small DBs, and dev environments.
Compute Optimized (C)
High-performance processors. Best for batch processing, scientific modeling, video encoding.
High-performance processors. Best for batch processing, scientific modeling, video encoding.
Memory Optimized (R, X, z)
Fast performance for large datasets in memory. Best for high-perf DBs and real-time analytics.
Fast performance for large datasets in memory. Best for high-perf DBs and real-time analytics.
Storage Optimized (I, D, H)
High, sequential R/W for local storage. Best for NoSQL DBs, data warehousing, and Hadoop.
High, sequential R/W for local storage. Best for NoSQL DBs, data warehousing, and Hadoop.
Accelerated Computing (P, G, Inf)
Hardware accelerators (GPUs/FPGAs). Best for Machine Learning and Graphics rendering.
Hardware accelerators (GPUs/FPGAs). Best for Machine Learning and Graphics rendering.
Burstable (T2/T3)
Baseline performance with ability to burst using CPU credits.
Baseline performance with ability to burst using CPU credits.
EC2 Core Concepts
- AMI: Pre-configured OS and software templates.
- Security Groups: Stateful virtual firewalls (Allow rules only).
- Key Pairs: SSH/RDP access credentials.
- Instance Connect: Secure shell access via IAM without managing SSH keys.
Containers on AWS: ECS vs EKS vs Fargate
| Feature | Amazon ECS | Amazon EKS | AWS Fargate |
|---|---|---|---|
| Model | AWS-native orchestration | Managed Kubernetes | Serverless Compute Engine |
| Complexity | Low (Easier to set up) | High (K8s expertise needed) | Minimal (No infra management) |
| Control | Deep AWS Integration | Open-source standard/Portable | Abstracted Infrastructure |
| Use Case | AWS-centric applications | Multi-cloud/K8s workflows | Event-driven/Unpredictable loads |
AWS Lambda Limits
- Execution Timeout 15 Minutes
- Memory Allocation 128MB – 10GB
- Ephemeral Storage (/tmp) 512MB – 10GB
- Deployment Size (Zip) 50MB Direct / 250MB S3
- Concurrency 1,000 (Default)
- Payload (Sync) 6MB
Serverless Support Services
API Gateway: Entry point for REST/WebSocket APIs; handles security & scaling.
DynamoDB: NoSQL database; millisecond latency; scales automatically.
Step Functions: Visual workflows (State Machines) to coordinate Lambda functions.
SAM: Template-based framework (YAML) for deploying serverless stacks.
Scaling Strategies
Vertical Scaling: Upgrading the instance size (More RAM/CPU on one box).
Horizontal Scaling: Adding more instances (Auto Scaling Groups). Preferred for high availability.
Commitment-Based Pricing
Reserved Instances (RI)
- Standard: Up to 72% discount. Best for steady-state.
- Convertible: Change instance family/OS. 66% discount.
- Capacity Reservation: Guarantees availability in specific AZ.
Savings Plans
- Compute Savings Plan: Most flexible. Applies to EC2, Fargate, and Lambda regardless of region/family.
- EC2 Instance Savings Plan: Region/Family specific. Higher discount (72%).