Right-size with an Auto Scaling group of 4 baseline instances scaling to 28
A company is hosting an application on AWS for a project that will run for the next 3 years. The application consists of 20 Amazon EC2 On-Demand Instances that are registered in a target group for a Network Load Balancer (NLB). The instances are spread across two Availability Zones. The application is stateless and runs 24 hours a day, 7 days a week. The company receives reports from users who are experiencing slow responses from the application. Performance metrics show that the instances are at 10% CPU utilization during normal application use. However, the CPU utilization increases to 100% at busy times, which typically last for a few hours. The company needs a new architecture to resolve the problem of slow responses from the application. Which solution will meet these requirements MOST cost-effectively?
Community Votes
100% of anonymous learners picked answer D. Votes are pick records left by other test-takers — they are not the verified answer.
Community Insight
Ten percent average CPU across twenty instances means roughly two instances worth of work, so a baseline of four with scaling to 28 matches the actual load profile, and Reserved Instances for only the baseline converts the steady-state cost while the peak capacity is paid for only while it is needed.
A stateless application runs 24 hours a day on 20 EC2 On-Demand instances behind an NLB across two Availability Zones, with CPU at ten percent during normal use and at one hundred percent during busy periods lasting a few hours. Users report slow responses during those peaks, and the fix must be the most cost-effective.
Setting the minimum capacity to 20 and the desired capacity to 28 with Reserved Instances for 20. That holds twenty instances running at all times when the workload needs about two, so the company keeps paying for eighteen idle instances, which is the opposite of cost-effective. Using a Spot Fleet is also wrong because a stateless workload running continuously for three years cannot rely on reclaimable capacity.
Community Discussion (3 comments)
Comments & Corrections
No comments yet — spotted an error or have a note? Share it below.
Expert Analysis
Why the Answer Is Correct
The metric that drives the design is the ten percent average CPU across twenty instances, which means the steady workload is roughly the equivalent of two instances. Setting the Auto Scaling group minimum capacity to 4 gives a baseline with headroom for the normal pattern, and the maximum of 28 covers the observed peak, so the group scales out for the few busy hours and scales back in afterwards, which is what removes the slow responses. Because the application is stateless and runs continuously for three years, the baseline of four is a predictable commitment, so purchasing Reserved Instances for four converts that steady cost at a discount while the peak instances are paid for only during the hours they are needed. The group attaches to the existing target group, so the load balancer design is unchanged.Why the Other Options Are Wrong
A: Setting the minimum capacity to 20 and the desired capacity to 28 keeps twenty instances running permanently and buys Reserved Instances for all twenty, so the company pays the reserved rate for eighteen instances that are idle outside the few peak hours, which is strictly more expensive than sizing the baseline to the real load. B and C: A Spot Fleet relies on Spot capacity that AWS may reclaim, and this application is stateless, runs continuously, and has a three-year project horizon, so building the steady state on Spot would risk availability for the sake of a discount that Reserved Instances already provide safely. C additionally replaces the NLB with an Application Load Balancer, which is an unnecessary change to a working layer 4 design and changes the service semantics rather than the capacity.Community Comment Notes
The community voted 100 to 0 for D, and the top-voted comments reasoned directly from the metrics, that ten percent CPU during normal use means the minimum capacity can be far lower than 20, so a minimum of 4 is the minimum configuration and the most cost-effective, with the peak hours absorbed by scaling out to the maximum of 28.Official Reference
Related Analysis
Practice All SAP-C02 Questions
Access 85 questions with complete answers and detailed explanations.
View Full SAP-C02 Practice Test →