AWS Certified Solutions Architect Associate (SAA-C03) Domain 3

Design High-Performing Architectures

Official Exam Guide: SAA-C03 Exam Guide

---

Domain Overview

Domain Weight: 24% of the exam

This domain tests your ability to design high-performing storage, compute, database, and networking solutions.

---

Task 3.1: Determine high-performing storage solutions

Storage Types Comparison

Storage Type Selection:

Use Case	Storage Type	Service
Object storage	S3	S3, Glacier
Block storage	EBS	EBS volumes
File storage	EFS/FSx	EFS, FSx
Archive	Glacier	S3 Glacier
Temporary	Instance Store	EC2 instance store

Amazon S3 Performance

S3 Performance Optimization:

• Multipart Upload - Files >100MB (required >5GB)
• Transfer Acceleration - Edge locations for faster uploads
• Byte-Range Fetches - Parallel downloads
• S3 Select - Query data in place (reduce transfer)

Request Rate:

• 3,500 PUT/COPY/POST/DELETE per prefix per second
• 5,500 GET/HEAD per prefix per second
• Use prefixes to scale

Best Practices:

• Randomize key names for high request rates
• Use CloudFront for read-heavy workloads
• Enable Transfer Acceleration for global uploads
• Use multipart for large files

AWS Documentation:

• S3 Performance Optimization

Amazon EBS Performance

Volume Types:

SSD-based:

• gp3 - General purpose, 16,000 IOPS, 1,000 MB/s
• gp2 - General purpose, bursts to 3,000 IOPS
• io2/io1 - Provisioned IOPS, up to 64,000 IOPS
• io2 Block Express - Up to 256,000 IOPS

HDD-based:

• st1 - Throughput optimized, 500 MB/s
• sc1 - Cold HDD, 250 MB/s

Selection Criteria:

• Databases, boot volumes → gp3 or io2
• High IOPS requirements → io2
• Big data, data warehouses → st1
• Infrequent access → sc1

EBS Optimization:

• Enable EBS-optimized instances
• Use RAID 0 for increased IOPS
• Snapshots don’t impact performance

AWS Documentation:

• EBS Volume Types

Amazon EFS Performance

Performance Modes:

• General Purpose - Low latency (most use cases)
• Max I/O - High aggregate throughput (big data)

Throughput Modes:

• Bursting - Scales with size
• Provisioned - Fixed throughput regardless of size
• Elastic - Auto-scales (recommended)

Storage Classes:

• Standard - Frequently accessed
• Infrequent Access (IA) - Lower cost for less accessed files

AWS Documentation:

• EFS Performance

FSx Family

FSx for Windows File Server:

• Native Windows file system
• SMB protocol
• Active Directory integration
• Use case: Windows applications

FSx for Lustre:

• High-performance computing (HPC)
• Machine learning workloads
• Sub-millisecond latencies
• Integrates with S3

FSx for NetApp ONTAP:

• NetApp features on AWS
• Multi-protocol (NFS, SMB, iSCSI)

FSx for OpenZFS:

• OpenZFS file system
• NFS protocol

AWS Documentation:

• Amazon FSx

---

Task 3.2: Design high-performing compute solutions

EC2 Instance Types

Instance Families:

General Purpose (T, M):

• Balanced compute, memory, networking
• Use case: Web servers, small databases

Compute Optimized (C):

• High-performance processors
• Use case: Batch processing, gaming, HPC

Memory Optimized (R, X, z):

• Large memory
• Use case: In-memory databases, big data

Storage Optimized (I, D, H):

• High sequential read/write
• Use case: NoSQL databases, data warehousing

Accelerated Computing (P, G, F, Inf):

• GPU, FPGA
• Use case: ML, graphics rendering

AWS Documentation:

• EC2 Instance Types

EC2 Placement Groups

Cluster:

• Low latency, high throughput
• Same AZ
• Use case: HPC, tightly coupled applications

Spread:

• Each instance on different hardware
• Up to 7 instances per AZ
• Use case: Critical instances

Partition:

• Groups of instances on separate partitions
• Use case: Hadoop, Cassandra, Kafka

AWS Documentation:

• Placement Groups

Container Services

Amazon ECS:

• AWS-native container orchestration
• EC2 or Fargate launch types
• Use case: Docker containers on AWS

Amazon EKS:

• Managed Kubernetes
• Use case: Kubernetes workloads

AWS Fargate:

• Serverless compute for containers
• No EC2 management
• Use case: Simplified container deployment

AWS Documentation:

• Amazon ECS
• Amazon EKS

---

Task 3.3: Determine high-performing database solutions

Database Selection

Database Types:

Relational (SQL):

• RDS - Managed MySQL, PostgreSQL, Oracle, SQL Server, MariaDB
• Aurora - AWS high-performance database
• Use case: ACID transactions, structured data

NoSQL:

• DynamoDB - Key-value, document store
• DocumentDB - MongoDB-compatible
• Neptune - Graph database
• Use case: High scale, flexible schema

In-Memory:

• ElastiCache - Redis, Memcached
• Use case: Caching, session storage

Data Warehouse:

• Redshift - Petabyte-scale analytics
• Use case: BI, analytics

Time-Series:

• Timestream - Time-series data
• Use case: IoT, monitoring

RDS Performance

Read Replicas:

• Asynchronous replication
• Up to 15 replicas (Aurora)
• Cross-region support
• Read scaling, not HA

Performance Optimization:

• Use appropriate instance type
• Provision adequate IOPS
• Enable Performance Insights
• Use read replicas for read-heavy workloads
• Cache frequently accessed data

Aurora Performance:

• 5x faster than MySQL
• 3x faster than PostgreSQL
• Auto-scaling storage
• Fast cloning
• Aurora Serverless for variable workloads

AWS Documentation:

• RDS Read Replicas
• Amazon Aurora

DynamoDB Performance

Performance Features:

• Single-digit millisecond latency
• Auto-scaling
• DynamoDB Accelerator (DAX) - microsecond caching
• Global Tables - multi-region replication
• On-demand pricing - no capacity planning

Optimization:

• Use partition keys wisely
• Avoid hot partitions
• Use DAX for read-heavy workloads
• Use Global Secondary Indexes (GSI) for query flexibility

AWS Documentation:

• DynamoDB Performance

Redshift Performance

Performance Features:

• Columnar storage
• Data compression
• Massively parallel processing (MPP)
• Result caching

Optimization:

• Distribution styles (KEY, ALL, EVEN)
• Sort keys for query performance
• Vacuum and analyze regularly
• Concurrency scaling

AWS Documentation:

• Amazon Redshift

---

Task 3.4: Determine high-performing networking solutions

VPC Design

Subnet Design:

• Public subnets (internet-facing resources)
• Private subnets (internal resources)
• Use all available AZs
• Plan CIDR blocks for growth

NAT Gateway:

• Managed NAT service
• High availability (AZ-level)
• Better than NAT instance
• Place in public subnet

AWS Documentation:

• VPC Design

Connectivity Options

AWS Direct Connect:

• Dedicated network connection
• Consistent network performance
• Lower latency than internet
• Use case: Hybrid cloud, large data transfers

VPN:

• Encrypted connection over internet
• Quick to set up
• Lower cost than Direct Connect
• Use case: Backup connectivity, temporary needs

Transit Gateway:

• Hub-and-spoke network topology
• Connect VPCs and on-premises
• Simplifies complex network architectures

AWS Documentation:

• AWS Direct Connect
• AWS Transit Gateway

Content Delivery

CloudFront:

• Global CDN
• Edge caching
• Reduces origin load
• SSL/TLS termination
• Lambda@Edge for customization

Global Accelerator:

• Static anycast IPs
• Intelligent routing
• For non-HTTP/HTTPS (TCP/UDP)
• Health checks and failover

AWS Documentation:

• Amazon CloudFront

Enhanced Networking

Placement Groups:

• Cluster placement for low latency
• Enhanced networking (SR-IOV)
• Up to 100 Gbps

Elastic Fabric Adapter (EFA):

• Network interface for HPC
• OS-bypass for lower latency
• Use case: MPI applications, ML training

AWS Documentation:

• Enhanced Networking

---

Exam Tips

Common Patterns:

1. High IOPS database → io2 EBS or Aurora
2. Cache layer → ElastiCache (Redis/Memcached)
3. Static content delivery → CloudFront + S3
4. Read-heavy database → RDS read replicas
5. Low latency between instances → Cluster placement group
6. Large file uploads → S3 multipart upload
7. Global users → CloudFront or Global Accelerator
8. HPC workloads → Cluster placement + EFA

Performance Principles:

• Right-size instances and storage
• Use caching liberally
• Scale horizontally when possible
• Use managed services for auto-scaling
• Leverage edge locations (CloudFront)

---

Final Thoughts

Domain 3 (Design High-Performing Architectures) is 24% of the exam.

Master these concepts:

1. Storage types and when to use each
2. EBS volume types (gp3, io2, st1)
3. EC2 instance types and placement groups
4. Database selection (RDS, DynamoDB, Redshift)
5. CloudFront for content delivery
6. ElastiCache for caching
7. RDS read replicas vs Multi-AZ

Performance = Right Service + Right Configuration + Caching + Scaling!