# Elasticity

Elasticity is the ability to increase and decrease computing resources to match changes in workload demand.

Cloud concept. Updated 2026-10-29. Page: https://blowstack.com/tools/cloud-concepts-hub/elasticity

> Elasticity aligns available capacity with actual demand, adding resources during busy periods and releasing them when demand falls. In the cloud, automated provisioning and usage-based pricing can reduce idle capacity and avoid purchasing infrastructure for occasional peaks, while helping applications maintain performance.

Practice exams covering Elasticity: [ANS-C01 AWS Certified Advanced Networking - Specialty](https://blowstack.com/practice-exams/aws-certified-advanced-networking-specialty), [CLF-C02 AWS Certified Cloud Practitioner](https://blowstack.com/practice-exams/aws-certified-cloud-practitioner), [SOA-C03 AWS Certified CloudOps Engineer - Associate](https://blowstack.com/practice-exams/aws-certified-cloudops-engineer-associate), [DEA-C01 AWS Certified Data Engineer - Associate](https://blowstack.com/practice-exams/aws-certified-data-engineer-associate), [DOP-C02 AWS Certified DevOps Engineer - Professional](https://blowstack.com/practice-exams/aws-certified-devops-engineer-professional), [AIP-C01 AWS Certified Generative AI Developer - Professional](https://blowstack.com/practice-exams/aws-certified-generative-ai-developer-professional), [MLA-C02 (beta) AWS Certified Machine Learning Engineer - Associate](https://blowstack.com/practice-exams/aws-certified-machine-learning-engineer-associate), [SAA-C03 AWS Certified Solutions Architect - Associate](https://blowstack.com/practice-exams/aws-certified-solutions-architect-associate), [SAP-C03 AWS Certified Solutions Architect - Professional](https://blowstack.com/practice-exams/aws-certified-solutions-architect-professional).

## 6 key aspects of Elasticity

Elasticity combines demand-aligned capacity changes, safe automation, and resource release to balance performance and cost within operational limits.

1. **Capacity follows demand**: Increase resources when workloads grow and release them when demand falls. Both expansion and contraction are central to elasticity.
2. **Automated resource adjustment**: Use demand-based policies or scheduled actions to adjust capacity without repeated manual intervention. Choose metrics that reflect workload pressure, such as requests per target or queue backlog per worker.
3. **Reduced idle-capacity costs**: Releasing unused resources can reduce usage-based charges and the need to provision permanently for peak demand. Savings depend on pricing, minimum capacity, and charges for resources that remain allocated.
4. **Application readiness**: Design components so new capacity can accept work and removed capacity can shut down safely. Externalized session state and graceful connection draining help prevent scaling events from disrupting users.
5. **Bounded response time and capacity**: Account for provisioning delays, warm-up periods, quotas, and downstream limits. Set minimum and maximum capacity deliberately rather than assuming resources appear instantly or without limits.
6. **Separate from availability and agility**: Elasticity matches resources to demand; high availability addresses continued operation through failures; agility concerns the speed of experimentation and delivery. A sound architecture may need all three, but none automatically guarantees the others.

## Cloud services for Elasticity

For an elastic web application on AWS, Amazon CloudWatch supplies workload metrics, Amazon EC2 Auto Scaling adjusts instance capacity, and Elastic Load Balancing distributes requests across healthy targets. AWS Application Auto Scaling provides a similar policy-driven mechanism for supported resource dimensions, including container task counts and provisioned database throughput. AWS Lambda handles execution-capacity scaling for function-based workloads without a customer-managed server fleet. These mechanisms still require suitable application design, sensible limits, and monitoring: adding compute cannot remove every downstream bottleneck, and releasing compute does not eliminate all associated infrastructure charges.

- **[Amazon EC2 Auto Scaling](https://blowstack.com/tools/cloud-services-hub/amazon-ec2-auto-scaling)**: Adjusts the number of EC2 instances in an Auto Scaling group using scaling policies or scheduled actions. Minimum, desired, and maximum capacity settings define the operating range, while instance warm-up and lifecycle handling help manage transitions.
- **[AWS Lambda](https://blowstack.com/tools/cloud-services-hub/aws-lambda)**: Automatically scales function execution capacity in response to incoming work, subject to concurrency quotas, scaling limits, and event-source behavior. It illustrates elasticity without requiring engineers to manage a fleet of servers.
- **[Amazon CloudWatch](https://blowstack.com/tools/cloud-services-hub/amazon-cloudwatch)**: Provides metrics and alarms that can inform or trigger scaling actions. Monitoring workload pressure, latency, and errors helps engineers determine whether capacity changes are meeting performance goals.

## Elasticity exam questions

Real practice exam questions on Elasticity, with the correct answer and its explanation.

### Question 1 (CLF-C02)

**Question:**

A retailer has configured automatic scaling for an Amazon EC2 Auto Scaling group. After a promotional event, a manager wants evidence that compute capacity increased as demand rose and decreased after demand fell. The group's instance-count metrics are enabled. Which approach best demonstrates this behavior without changing the scaling configuration?

**Correct answer:** Use an Amazon CloudWatch dashboard to compare application request volume with the group's in-service instance count throughout the promotion and afterward.

**Explanation:** CloudWatch can display demand metrics alongside Auto Scaling group instance-count metrics over time. This comparison provides evidence that capacity increased during high demand and was released when demand fell, demonstrating elasticity.

### Question 2 (MLA-C02 (beta))

**Question:**

A company hosts three CPU-based NLP models on separate Amazon SageMaker AI real-time endpoints. The models use different serving containers, and their traffic peaks occur at different times. Monitoring shows substantial unused memory and CPU capacity across the endpoints.

The company wants to consolidate the models onto a shared pool of endpoint instances. Each model must remain independently addressable, and the number of running serving copies of each model must scale independently with its request volume. The underlying instance pool must also grow and shrink as the combined capacity requirement changes.

Which solution best meets these requirements?

**Correct answer:** Create an inference component for each model on a shared endpoint, specifying its compute and memory requirements. Configure automatic scaling of each component's copy count and enable managed instance scaling for the endpoint.

**Explanation:** Inference components allow independently addressable models and their containers to share endpoint instance resources. Resource requirements guide placement, each component's copy count can scale independently, and managed instance scaling adjusts the underlying instance pool to accommodate the components' capacity needs.

### Question 3 (SAP-C03)

**Question:**

A company runs a Linux container worker service on Amazon ECS with AWS Fargate. An existing, validated scaling policy adjusts the service between 6 and 60 tasks based on workload demand. Each task currently requests 2 vCPUs and 16 GiB of memory.

Metrics collected across a complete business cycle show that each task requires at most 0.7 vCPU and 6.2 GiB of memory. Representative load tests confirm that tasks configured with 1 vCPU and 8 GiB of memory maintain the required processing throughput.

The company requires six tasks to run on capacity that is not subject to Spot interruptions. Additional tasks process checkpointed work that can tolerate interruptions and can wait when discounted capacity is unavailable. The company wants to minimize compute charges without changing the scaling policy or managing servers.

Which solution best meets these requirements?

**Correct answer:** Deploy a task definition with 1 vCPU and 8 GiB of memory. Configure the service capacity provider strategy with FARGATE base 6 and weight 0, and FARGATE_SPOT weight 1.

**Explanation:** The tested task size reduces billed CPU and memory allocations while preserving throughput. The base assigns the first six tasks to FARGATE. After the base is satisfied, the weights direct all additional tasks to FARGATE_SPOT. This preserves the non-Spot baseline and uses discounted, interruptible capacity for work that can tolerate interruptions and capacity shortages.

### Question 4 (SOA-C03)

**Question:**

A company runs queue-processing workers as an Amazon ECS service on AWS Fargate. Each task processes one job at a time, and each job takes up to 25 minutes. ECS Service Auto Scaling adjusts the desired task count based on a custom workload metric.

During scale-in, ECS sometimes stops tasks that are still processing jobs. The company must allow active jobs to finish while continuing to remove idle tasks automatically. Worker code changes are permitted, but the company does not want to build a custom controller to select tasks for termination.

Which solution best meets these requirements?

**Correct answer:** Configure the workers to enable ECS task scale-in protection before polling for a job, with a 60-minute expiration. Clear protection when no job is available or after processing finishes.

**Explanation:** ECS task scale-in protection prevents protected service tasks from being terminated during service scale-in. Enabling protection before acquiring work avoids leaving an active job unprotected. A 60-minute expiration exceeds the maximum job duration, and clearing protection when idle allows ECS to remove tasks without a custom termination controller.

### Question 5 (AIP-C01)

**Question:**

A company hosts a foundation model on an Amazon SageMaker AI real-time endpoint. Traffic rises and falls gradually throughout the day, with total outstanding requests varying from 8 to 60. Request duration varies significantly because generated answers have different lengths. The company currently keeps 12 GPU instances running continuously.

Load tests show that targeting six outstanding requests per instance provides sufficient headroom to meet the latency objective. Outstanding requests include requests being processed and requests waiting for inference. New instances become ready within 90 seconds, and observed traffic changes slowly enough for reactive scaling to meet the latency objective. Required quotas and GPU capacity are available.

GPU memory utilization remains near 90% even when the endpoint is idle because the model stays loaded. CPU utilization has little correlation with inference demand. The company requires at least two instances to remain available.

Which configuration should a generative AI developer implement to reduce idle GPU costs while maintaining the latency objective?

**Correct answer:** Configure Application Auto Scaling with a minimum of 2 instances and a maximum of 12. Use target tracking with a target of six outstanding requests per instance, publishing a custom metric that divides total outstanding requests by the number of in-service instances. Keep scale-in enabled.

**Explanation:** Outstanding requests per instance measures demand relative to available serving capacity. Adding instances reduces the metric, and declining demand allows the policy to remove instances down to the required minimum. The target is supported by load testing, and the stated startup time and traffic ramp make reactive scaling appropriate.

### Question 6 (DEA-C01)

**Question:**

A company runs operational analytics on a provisioned Amazon Redshift cluster. Queries arrive throughout the business day, but unpredictable 10-minute bursts require several times the normal query-processing capacity. The current cluster is sized for normal demand, so queries queue during these bursts.

The company is willing to migrate the warehouse. A data engineer must select a configuration that automatically increases compute capacity during bursts and reduces it afterward, without custom scaling automation. The normal operating capacity must not be sized for peak demand. Testing has established the compute capacity needed for both normal and peak workloads, and the necessary service quotas are available.

Which solution meets these requirements?

**Correct answer:** Migrate to Amazon Redshift Serverless. Set base capacity for normal demand and maximum capacity high enough to accommodate the tested peak workload.

**Explanation:** Redshift Serverless can automatically increase compute capacity as workload demand rises and reduce it as demand falls. A normal-demand base capacity and a sufficiently high maximum capacity allow this elasticity without sizing the normal operating capacity for peaks or implementing custom scaling automation.

### Question 7 (DOP-C02)

**Question:**

A company runs an Amazon ECS worker service on On-Demand EC2 instances through an Auto Scaling group capacity provider. Capacity provider managed scaling and managed termination protection are enabled, and the Auto Scaling group has instance scale-in protection configured correctly. ECS Service Auto Scaling adjusts the service's desired task count according to queue demand.

Each task processes one job at a time. Jobs take 20–90 minutes, cannot be checkpointed, and can start at any time. During decreases in the service's desired count, ECS sometimes stops tasks that are still processing jobs, even though their EC2 instances remain running. The company also needs to deploy new task revisions without interrupting active jobs.

A DevOps engineer must protect active jobs during service scale-in and rolling deployments while allowing idle tasks and unused EC2 capacity to be removed automatically. The worker application can be modified, but the company does not want a separate orchestration service.

Which solution meets these requirements?

**Correct answer:** Have each worker enable ECS task scale-in protection through the ECS agent endpoint before processing a job, with a 120-minute expiration. Clear protection when the job finishes, and renew it before expiration if necessary. Retain capacity provider managed scaling and managed termination protection.

**Explanation:** ECS task scale-in protection prevents protected service tasks from being stopped by service scale-in or rolling deployments. Setting protection before processing and clearing it afterward ties protection to the actual job lifecycle. The expiration exceeds the stated maximum job duration, and renewal accommodates unexpected overruns. Existing capacity provider settings continue to manage EC2 capacity and protect instances that host non-daemon tasks during infrastructure scale-in.

### Question 8 (SAA-C03)

**Question:**

A company operates a document-management application with two storage components. An Amazon RDS for PostgreSQL DB instance stores document metadata in 300 GiB of allocated General Purpose SSD storage. An Amazon EFS Regional file system stores the documents themselves. Both datasets grow gradually but unpredictably. During the next year, metadata could require 1.5 TiB, and documents could require 20 TiB. Growth is slow enough to allow automatic storage modifications to complete before space is exhausted.

Testing confirms that the current database compute capacity and file system throughput will remain sufficient. The company wants both components to accommodate storage growth without operator intervention while minimizing charges for unused capacity and avoiding custom automation.

Which solution should a solutions architect recommend?

**Correct answer:** Enable RDS storage autoscaling with a maximum storage threshold of 2 TiB. Keep the existing EFS file system without adding capacity-scaling automation.

**Explanation:** RDS storage autoscaling increases allocated database storage as needed, up to the configured maximum, avoiding an immediate allocation for projected peak demand. EFS automatically grows as files are added and does not require provisioned storage capacity or capacity-scaling policies. Its existing throughput is sufficient.

### Question 9 (ANS-C01)

**Question:**

A company runs a ticket-sales application behind an Application Load Balancer (ALB) across three Availability Zones. Once a month, traffic increases from its normal level to a much higher, benchmarked load within seconds. Load tests show that reactive scaling of the ALB can lag behind this increase. The application targets are provisioned in advance and have sufficient capacity, and the ALB subnets have sufficient available IP addresses.

The company must verify that the required load balancer capacity is available before ticket sales begin. After each event, load balancer capacity must be allowed to decrease with demand without retaining a paid event-sized capacity floor. Generating synthetic traffic to prepare the ALB is prohibited.

Which approach meets these requirements?

**Correct answer:** Configure an ALB minimum capacity reservation based on the benchmarked event demand. Verify that the reservation is provisioned before sales begin, and remove the reservation after the event so that capacity can again decrease without the event-sized minimum.

**Explanation:** An ALB capacity reservation establishes a minimum capacity in LCUs that can be prepared before a sudden increase in demand. The ALB can still scale above that minimum if needed. Removing the reservation after the event removes the paid capacity floor and allows normal demand-driven scaling, including scaling down.

## Practice Elasticity in a full exam

Questions on Elasticity in each practice exam, with every answer option and an explanation for each:

- [SAA-C03](https://blowstack.com/practice-exams/aws-certified-solutions-architect-associate): 26 questions
- [CLF-C02](https://blowstack.com/practice-exams/aws-certified-cloud-practitioner): 15 questions
- [SOA-C03](https://blowstack.com/practice-exams/aws-certified-cloudops-engineer-associate): 11 questions
- [MLA-C02 (beta)](https://blowstack.com/practice-exams/aws-certified-machine-learning-engineer-associate): 9 questions
- [DOP-C02](https://blowstack.com/practice-exams/aws-certified-devops-engineer-professional): 9 questions
- [SAP-C03](https://blowstack.com/practice-exams/aws-certified-solutions-architect-professional): 3 questions
- [AIP-C01](https://blowstack.com/practice-exams/aws-certified-generative-ai-developer-professional): 3 questions
- [DEA-C01](https://blowstack.com/practice-exams/aws-certified-data-engineer-associate): 1 question
- [ANS-C01](https://blowstack.com/practice-exams/aws-certified-advanced-networking-specialty): 1 question

## What is Elasticity

Elasticity means adjusting resource capacity as demand changes, in both directions. An elastic application can acquire additional capacity during a traffic spike and release that capacity afterward. The goal is not simply to make a system larger; it is to keep capacity appropriately aligned with the work being performed. Cloud platforms make this practical by providing resources through APIs instead of requiring engineers to purchase, install, and later retire physical equipment.

Consider an online store that normally needs two application servers but requires twelve during a sale. An elastic design adds servers as demand rises and removes the additional servers when traffic returns to normal. Without elasticity, the business might keep twelve servers running continuously or accept poor performance during the sale. Releasing unnecessary resources can reduce usage charges, although retained storage, minimum capacity, and other supporting infrastructure may still incur costs.

Elasticity and scalability are related but different. Scalability describes a system's ability to handle changing workload levels by changing capacity. Elasticity emphasizes matching capacity to fluctuating demand and relinquishing resources when they are no longer needed. Horizontal scaling adds or removes instances, containers, or workers; vertical scaling changes the resources assigned to an individual component. Both can support elasticity, but vertical changes may require restarts and have tighter size limits.

Automatic scaling is a common mechanism for implementing elasticity. Policies can respond to CPU utilization, request volume, or queue backlog, while scheduled actions can prepare capacity for known events. For example, background workers might scale out when queued jobs accumulate and scale in after processing catches up. On AWS, Amazon EC2 Auto Scaling provides elasticity by adjusting the number of EC2 instances according to configured policies and capacity limits. Provisioning a fixed set of cloud instances does not automatically make an application elastic.

Application design determines whether additional resources are useful. Stateless request handlers, shared session storage, distributed work queues, and graceful shutdown procedures make it easier to add and remove workers safely. A database bottleneck or an external API limit can prevent extra application servers from improving throughput. Scaling signals should therefore reflect the actual constraint rather than relying on CPU usage alone.

Elasticity is not instantaneous or unlimited. Resource quotas, available capacity, instance startup time, application initialization, and scaling-policy settings affect response speed. Minimum capacity can protect latency during sudden bursts, while maximum capacity limits spending and resource consumption. Scale-in must allow active requests or jobs to finish, and poorly tuned policies can repeatedly add and remove resources.

Elasticity is also distinct from high availability and agility. High availability keeps a workload accessible despite failures, typically through redundancy and fault isolation. Agility is the ability to experiment and deliver changes quickly. An elastic workload can still have a single point of failure, and a highly available workload can use fixed capacity. For exam questions, associate elasticity with matching resources to changing demand, high availability with continued operation, and agility with faster experimentation and delivery.

## Learn Elasticity in lessons

21 short lessons on Elasticity in 8 certification courses.

- CLF-C02: Elasticity: matching capacity to demand (Premium), https://app.blowstack.com/lessons/elasticity-matching-capacity-to-demand-clf-c02
- MLA-C02 (beta): Elasticity vs. horizontal scaling (Premium), https://app.blowstack.com/lessons/elasticity-vs-horizontal-scaling-mla-c02
- SAP-C03: Elasticity: tuning capacity response to demand (Premium), https://app.blowstack.com/lessons/elasticity-tuning-capacity-response-to-demand-sap-c03
- SOA-C03: Elasticity: demand-responsive capacity control (Premium), https://app.blowstack.com/lessons/elasticity-demand-responsive-capacity-control-soa-c03
- AIP-C01: Capacity Planning vs Elasticity (Premium), https://app.blowstack.com/lessons/capacity-planning-vs-elasticity-aip-c01
- DEA-C01: Elasticity: capacity adjustment in provisioned and serverless services (Premium), https://app.blowstack.com/lessons/elasticity-capacity-adjustment-in-provisioned-and-serverless-services-dea-c01
