SAA-C03 Design Cost-Optimized Architectures Practice Question

A company runs a real-time bidding platform on Amazon EC2 instances that must respond within milliseconds. The workload is highly variable, with unpredictable spikes during business hours. The company wants to minimize costs while ensuring the application always has enough capacity to handle sudden traffic surges. Which pricing model should they use?

⚠ Common exam trap

The trap here is assuming that any commitment-based discount (Reserved Instances, Savings Plans) is always cheaper, but for highly variable demand, the flexibility of On-Demand with Auto Scaling often results in lower overall cost and better performance.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Use On-Demand Instances with an Auto Scaling group.

For unpredictable, latency-sensitive workloads, On-Demand Instances combined with Auto Scaling provide the necessary elasticity and cost control. You pay only for what you use, and Auto Scaling ensures capacity matches demand. Reserved or Savings Plans are better for steady-state usage, while Spot Instances introduce interruption risk. Dedicated Hosts are cost-prohibitive and inflexible for this scenario.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✗
Use a Savings Plan with a 3-year term to cover all compute usage.
Why it's wrong here
Savings Plans offer discounts in exchange for a commitment to a consistent amount of compute usage (measured in $/hour) over a term. However, they do not provide the elasticity needed for unpredictable spikes; you would still need On-Demand or Spot capacity for excess usage. Committing to a 3-year term for a variable workload could result in paying for unused commitment during low-traffic periods, increasing costs rather than reducing them.
✗
Purchase a 1-year All Upfront Reserved Instance for the baseline capacity and use Spot Instances for spikes.
Why it's wrong here
Reserved Instances are cost-effective for steady-state usage, but the bidding platform has highly variable demand. Spikes cannot be reliably handled by Spot Instances because they can be interrupted with a two-minute warning, causing missed bids. This combination would risk performance and potentially cost more due to unused reserved capacity during low periods. The scenario requires immediate, uninterrupted scaling, which this model does not guarantee.
✓
Use On-Demand Instances with an Auto Scaling group.
Why this is correct
On-Demand Instances provide the flexibility to scale up and down without long-term commitment, matching the unpredictable spikes. An Auto Scaling group automatically adjusts capacity to maintain performance. While Spot Instances could be cheaper, they risk interruption, which is unacceptable for a latency-sensitive bidding platform. Savings Plans or Reserved Instances would not provide the needed elasticity for unpredictable surges and could lead to overprovisioning or insufficient capacity.
✗
Use Dedicated Hosts to ensure performance and reduce costs.
Why it's wrong here
Dedicated Hosts provide physical servers dedicated to your use, which is beneficial for licensing and compliance but significantly more expensive than shared tenancy. They do not offer cost savings for variable workloads, as you pay for the entire host even when utilization is low. Additionally, they do not automatically scale with demand, making them unsuitable for a highly variable, latency-sensitive application.

Go deeper

SAA-C03 Design Cost-Optimized Architectures Practice Question

Why each option matters

Option-by-option breakdown

Related to this question