Task Statement 4.3: Design cost-optimized database solutions.

📘AWS Certified Solutions Architect – (SAA-C03)

1. What is Database Capacity Planning?

Database capacity planning is the process of deciding:

How much read and write performance your database needs
How much storage is required
How the database will scale over time
How to meet performance needs at the lowest possible cost

In AWS, capacity planning is important because many database services use capacity-based pricing models.

2. What are Capacity Units?

Capacity Units are a way AWS measures how much work a database can perform.

They define:

How many reads per second are allowed
How many writes per second are allowed

Different AWS database services use different types of capacity units.

3. Capacity Units in Amazon DynamoDB

3.1 Read Capacity Units (RCUs)

A Read Capacity Unit (RCU) represents:

1 strongly consistent read per second for an item up to 4 KB

2 eventually consistent reads per second for an item up to 4 KB

Key Points:

Strong consistency = latest data
Eventual consistency = slightly delayed but cheaper
Larger items consume more RCUs

3.2 Write Capacity Units (WCUs)

A Write Capacity Unit (WCU) represents:

1 write per second for an item up to 1 KB

Key Points:

Larger items consume more WCUs
Writes are always strongly consistent

3.3 Example (Important for Exam)

If:

Item size = 2 KB
Writes = 100 per second

Then:

Each write consumes 2 WCUs
Total required = 200 WCUs

4. DynamoDB Capacity Modes

4.1 Provisioned Capacity Mode

You manually define:

Number of RCUs
Number of WCUs

Benefits:

Lower cost if workload is predictable
Full control

Risks:

Over-provisioning → wasted money
Under-provisioning → throttling

4.2 On-Demand Capacity Mode

No need to define RCUs/WCUs
AWS automatically scales capacity

Benefits:

Best for unpredictable workloads
No planning required

Drawback:

Higher cost compared to provisioned (if workload is stable)

5. Auto Scaling in DynamoDB

Auto Scaling adjusts RCUs/WCUs automatically based on usage.

How it works:

Define minimum and maximum limits
Set a target utilization (e.g., 70%)

Benefits:

Prevents over-provisioning
Avoids throttling
Optimizes cost

6. Burst Capacity (Important Concept)

DynamoDB allows temporary bursts above provisioned capacity.

Key Points:

Uses unused capacity from previous time
Helps handle sudden traffic spikes
Not guaranteed for long periods

7. Partitioning and Capacity

DynamoDB stores data in partitions.

Important limits per partition:

~3,000 RCUs
~1,000 WCUs

Why this matters:

Even if total capacity is high, poor partition design can cause hot partitions
Leads to throttling

8. Capacity Planning for Amazon RDS

Unlike DynamoDB, Amazon RDS does NOT use RCUs/WCUs.

Instead, capacity planning includes:

8.1 Compute Capacity

Instance type (CPU, RAM)
Example: db.t3, db.m5

8.2 Storage Capacity

Allocated storage (GB)
Storage types:
- General Purpose (gp2, gp3)
- Provisioned IOPS (io1, io2)

8.3 IOPS (Input/Output Operations Per Second)

Measures disk performance
Important for high-performance workloads

9. Capacity Planning for Amazon Aurora

9.1 Aurora Capacity Units (ACUs) – Serverless

In Aurora Serverless:

Uses Aurora Capacity Units (ACUs)

1 ACU includes:

Combination of CPU + memory + networking

Key Features:

Automatically scales up/down
Pay only for what you use

10. Key Differences Across Services

Service	Capacity Unit Type	Scaling Model
DynamoDB	RCUs / WCUs	Manual / Auto / On-Demand
RDS	Instance + IOPS	Vertical scaling
Aurora	ACUs	Automatic scaling

11. Cost Optimization Strategies (Exam Focus)

11.1 Right-Sizing Capacity

Avoid over-provisioning
Monitor usage with CloudWatch

11.2 Use On-Demand When:

Traffic is unpredictable
Application is new

11.3 Use Provisioned When:

Traffic is stable
Predictable workload

11.4 Enable Auto Scaling

Keeps cost low while maintaining performance

11.5 Optimize Item Size

Smaller items = fewer RCUs/WCUs

12. Monitoring Capacity

Use Amazon CloudWatch to track:

Consumed RCUs/WCUs
Throttled requests
CPU utilization (RDS)
Storage usage

13. Common Exam Scenarios

Scenario 1:

Application has unpredictable traffic
👉 Use DynamoDB On-Demand

Scenario 2:

Stable workload with predictable usage
👉 Use Provisioned capacity + Auto Scaling

Scenario 3:

Database experiencing throttling
👉 Increase RCUs/WCUs OR improve partition design

Scenario 4:

Need automatic scaling without managing instances
👉 Use Aurora Serverless (ACUs)

14. Key Exam Tips (Very Important)

RCU = reads, WCU = writes
4 KB read unit, 1 KB write unit
Eventually consistent reads cost less
On-Demand = flexible but more expensive
Provisioned = cheaper but requires planning
Auto Scaling = best balance
Hot partitions can cause throttling even with high capacity
Aurora Serverless uses ACUs, not RCUs/WCUs

15. Final Summary

Database capacity planning in AWS is about:

Matching performance needs with capacity units
Choosing the right scaling model
Avoiding over-provisioning and throttling
Using AWS features like Auto Scaling and On-Demand

Understanding how RCUs, WCUs, and ACUs work is essential to:

Design efficient systems
Reduce cost
Pass the SAA-C03 exam