Task Statement 3.2: Design high-performing and elastic compute solutions.
📘AWS Certified Solutions Architect – (SAA-C03)
When designing high-performing and elastic compute solutions in AWS, one of your key responsibilities is to choose the right type of compute resources that match the business requirements, workload patterns, and cost considerations. In AWS, compute options are primarily based on Amazon EC2 instances, but also include serverless options like Lambda and containers like Fargate.
1. Understanding EC2 Instance Types
Amazon EC2 provides many instance types. Each instance type is optimized for certain workloads. AWS groups them into families. Choosing the right instance type is crucial because it affects performance, cost, and scalability.
Main EC2 instance families:
| Family | Purpose / Features | Example Workloads |
|---|---|---|
| General Purpose (T, M series) | Balanced CPU, memory, network | Web servers, small databases, dev/test environments |
| Compute Optimized (C series) | High CPU relative to memory | High-performance web servers, batch processing, scientific modeling |
| Memory Optimized (R, X, Z series) | High memory relative to CPU | Large databases, in-memory caches, real-time analytics |
| Storage Optimized (I, D, H series) | High disk throughput and IOPS | Big data, data warehouses, transactional databases |
| Accelerated Computing (P, G, F series) | GPU or FPGA for specialized computing | Machine learning training/inference, video rendering, simulation |
Key points for the exam:
- T instances: Burstable CPU performance. Good for workloads that are not consistently CPU-intensive.
- M instances: Balanced CPU/memory. General-purpose, can handle steady workloads.
- C instances: CPU-heavy workloads. Compute-intensive applications.
- R/X/Z instances: For memory-intensive workloads. Think large in-memory databases.
- I/D/H instances: Storage-optimized for high disk IOPS or throughput.
- P/G/F instances: GPU/FPGA workloads for graphics or AI.
2. Selecting the Right EC2 Features
AWS EC2 instances have additional features that impact performance, scalability, and availability:
- EC2 Pricing Options:
- On-Demand: Pay per hour/second, flexible, no commitment.
- Reserved Instances: Lower cost for a 1- or 3-year commitment, good for steady workloads.
- Savings Plans: Flexible savings across instance families based on usage.
- Spot Instances: Very low cost, for interruptible workloads like batch jobs.
- Storage Options:
- EBS (Elastic Block Store): Persistent block storage attached to instances. You can choose:
- SSD for high IOPS (fast database)
- HDD for throughput-intensive workloads (big logs)
- Instance Store: Temporary local storage (data disappears if instance stops).
- EBS (Elastic Block Store): Persistent block storage attached to instances. You can choose:
- Networking Options:
- EC2 supports Elastic Network Interfaces (ENI) for multiple IPs or network isolation.
- Enhanced networking (using Elastic Network Adapter) for high throughput and low latency.
- Placement Options:
- Placement Groups: Control how instances are physically located for low latency or high availability.
- Cluster: Instances physically close for HPC (high-performance computing)
- Spread: Instances spread across hardware to reduce single points of failure
- Partition: Divides instances into logical partitions for fault tolerance
- Placement Groups: Control how instances are physically located for low latency or high availability.
- Elasticity Features:
- Auto Scaling: Dynamically adds/removes instances based on load.
- Load Balancers (ALB/NLB): Distribute traffic to multiple instances.
3. Other AWS Compute Options Beyond EC2
AWS provides alternatives when EC2 isn’t the best choice:
- AWS Lambda (Serverless):
- No servers to manage.
- Automatically scales with workload.
- Pay only for execution time (milliseconds).
- Ideal for event-driven workloads, such as processing files in S3 or responding to API requests.
- Amazon ECS / EKS with Fargate (Containers):
- Containers provide consistent environments.
- Fargate lets you run containers without managing servers.
- Good for microservices and batch processing.
- AWS Batch:
- Managed batch computing service.
- Automatically provisions the compute needed for jobs.
- Ideal for large-scale parallel processing.
4. Key Exam Tips for Selecting Compute Options
- Match the workload to the instance family: CPU-heavy → C series; memory-heavy → R/X series; balanced → M series.
- Consider cost vs. availability: Spot for cheap and interruptible workloads, Reserved/Savings Plan for predictable workloads.
- Check scaling needs: Use Auto Scaling and Elastic Load Balancing for variable traffic.
- Know alternative compute services: Lambda, Fargate, and Batch for serverless or containerized workloads.
- Understand storage and networking: EBS type, network performance, and placement groups can impact performance significantly.
5. Summary Table for Quick Exam Reference
| Requirement | AWS Option / Feature |
|---|---|
| Variable traffic | Auto Scaling + Load Balancer |
| CPU-intensive | C series EC2 |
| Memory-intensive | R/X series EC2 |
| Storage-heavy | I/D/H series EC2 |
| GPU / AI | P/G/F series EC2 |
| Short, event-driven tasks | Lambda |
| Containers without managing servers | Fargate |
| Large-scale batch jobs | AWS Batch |
| Cost optimization | Spot, Reserved, or Savings Plans |
| High availability | Placement Groups, Multi-AZ deployment |
✅ Exam tip: AWS may ask scenario-based questions where you must pick the instance type or compute option based on workload description. Focus on matching instance families and features to the workload characteristics.
