Question

Difficulty: EasyCost-Optimized Database Design and Capacity Planning

An educational institution runs a student voting application during the annual student council election. The application experiences a sudden, unpredictable surge in database writes for a single day, but remains completely idle for the rest of the year. The system uses an Amazon DynamoDB table. Which capacity mode should the solutions architect choose to minimize costs while preventing write throttling?

  1. On-Demand capacity modeAnswer
  2. B
    Provisioned capacity mode with Auto Scaling enabled
  3. C
    Provisioned capacity mode set to a high fixed capacity baseline
  4. D
    Provisioned capacity mode with a Compute Savings Plan applied

Answer

On-Demand capacity mode
On-Demand capacity mode is ideal for highly unpredictable and spiky workloads with long periods of inactivity. It provides instant scaling to handle traffic bursts without throttling and incurs zero cost when the database is idle, satisfying both performance and cost constraints.

Step-by-Step Solution

1
Analyze the database workload characteristics
The workload consists of a sudden, unpredictable peak lasting one day and a long idle period of 364 days.
Understanding the traffic pattern is critical for choosing between provisioned and on-demand database capacity models.
2
Evaluate the capacity scaling requirements
The scaling must be instantaneous to avoid throttling during the sudden spike, ruling out reactive auto-scaling models.
DynamoDB Auto Scaling relies on CloudWatch metrics and takes minutes to provision new throughput, leading to initial throttling on instant spikes.
3
Select the most cost-effective configuration that prevents throttling
On-Demand capacity mode provides pay-per-request pricing, scales instantly, and costs nothing when the table is idle.
This configuration aligns perfectly with the goal of minimizing idle costs while accommodating unpredictable, instantaneous bursts of traffic.

Key Concept

DynamoDB capacity planning for unpredictable and sporadic workloads
Estimated Time:45s
Rate this question