# Fault Isolation

Fault isolation is the design principle of containing failures within defined boundaries so they do not disrupt otherwise healthy parts of a system.

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

> Cloud systems use fault isolation to limit the scope of infrastructure failures, software defects, and resource exhaustion. Separating components into independent failure domains reduces correlated outages, but applications also need healthy alternatives and recovery mechanisms to remain available.

Practice exams covering Fault Isolation: [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), [DVA-C02 AWS Certified Developer - Associate](https://blowstack.com/practice-exams/aws-certified-developer-associate), [DOP-C02 AWS Certified DevOps Engineer - Professional](https://blowstack.com/practice-exams/aws-certified-devops-engineer-professional), [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).

## 5 key aspects of Fault Isolation

Effective fault isolation combines independent failure domains, containment controls, dependency-aware design, and tested recovery paths.

1. **Explicit failure domains**: Identify which resources can fail together because they share a host, facility, zone, dependency, or operational change. Choose deployment boundaries that match the failures the workload must survive.
2. **Independent infrastructure boundaries**: Availability Zones are physically separated and designed not to share infrastructure single points of failure. Place redundant components across zones when protection against a zone outage is required.
3. **Limited failure propagation**: Use separate resource pools, bounded queues, timeouts, and circuit breakers to keep a localized problem from exhausting healthy components. Apply these controls to software dependencies as well as infrastructure.
4. **End-to-end dependency separation**: Trace compute, data, networking, and supporting services across the full request path. A shared dependency in one failure domain can defeat otherwise well-distributed application servers.
5. **Redundancy and tested recovery**: Combine isolation with healthy alternatives, traffic redirection, and sufficient surviving capacity. Validate failure behavior through controlled tests, including database reconnection and recovery under increased load.

## Cloud services for Fault Isolation

EC2 supplies application instances placed across Availability Zones, while EC2 Auto Scaling manages their distribution and replacement. Elastic Load Balancing uses health checks to route requests to healthy application targets. RDS Multi-AZ extends protection to the database tier through cross-zone deployment and managed failover. Together, these services help a workload use infrastructure isolation boundaries, but they do not remove every shared dependency. Engineers must also verify outbound connectivity, session handling, database reconnection, and surviving capacity so that a zone failure does not become an application-wide outage.

- **[Amazon EC2](https://blowstack.com/tools/cloud-services-hub/amazon-ec2)**: EC2 lets engineers place instances in different Availability Zones to separate compute resources across infrastructure failure domains. Multiple instances in a single zone do not provide protection against loss of that zone.
- **[Amazon RDS](https://blowstack.com/tools/cloud-services-hub/amazon-rds)**: RDS Multi-AZ deployments maintain database resources across Availability Zones and support automatic failover for supported configurations. Applications must still handle interrupted connections and reconnect after failover; a Single-AZ database remains a zonal dependency.
- **[Amazon EC2 Auto Scaling](https://blowstack.com/tools/cloud-services-hub/amazon-ec2-auto-scaling)**: An Auto Scaling group configured across multiple Availability Zones distributes instances and attempts to maintain desired capacity as conditions change. Maintain sufficient surviving capacity rather than assuming replacement instances will become available immediately during an outage.

## Fault Isolation exam questions

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

### Question 1 (SOA-C03)

**Question:**

A company runs application instances in private subnets in Availability Zones A and B. The applications must call the regional AWS Security Token Service (AWS STS) API to obtain credentials. The VPC has no NAT gateway or other internet access path. An AWS STS interface VPC endpoint has private DNS enabled and an endpoint network interface only in Zone A.

During a Zone A outage, applications in Zone B remained healthy but could not refresh their credentials. The endpoint security group and network ACLs already permit the required traffic from both application subnets. The applications use the standard regional AWS STS hostname and retry failed requests with fresh DNS lookups.

Which change should a CloudOps engineer make so that applications in a surviving zone can continue obtaining credentials after either zone becomes unavailable?

**Correct answer:** Modify the existing AWS STS interface endpoint to include a subnet in Zone B while retaining its subnet in Zone A and private DNS configuration.

**Explanation:** Adding a subnet in Zone B creates an endpoint network interface there. The endpoint then provides private connectivity through network interfaces in both zones. With private DNS and application DNS refresh and retries, applications can use a surviving endpoint interface without depending exclusively on Zone A.

### Question 2 (SAP-C03)

**Question:**

A logistics company processes shipment updates for 20 external partners. Updates enter one Amazon SQS Standard queue. An Amazon EC2 Auto Scaling group spans three Availability Zones, and every worker can process updates for any partner. Processing an update requires a synchronous call to that partner's API.

During an incident, one partner's API began responding very slowly. Updates for that partner occupied nearly all worker execution slots, and its growing backlog delayed updates for healthy partners. The workers remained healthy. Adding instances helped temporarily, but the affected partner eventually consumed the additional capacity.

The company requires healthy partners to continue meeting their processing latency targets throughout another partner's prolonged API degradation. Updates must remain queued for eventual processing during outages of up to 24 hours. Each partner's normal peak workload is known, and sufficient capacity can be provisioned in advance. The ingestion tier and other dependencies have sufficient capacity and are not bottlenecks.

Which architecture change meets these requirements?

**Correct answer:** Create one SQS queue and one dedicated Auto Scaling group for each partner. Route updates to the appropriate queue, and restrict each group to consuming only its partner's queue. Preprovision each group for that partner's normal peak and configure queue retention to exceed the outage and recovery period.

**Explanation:** Separating both queues and worker fleets contains the degraded partner's backlog and occupied execution slots. Healthy partners retain their own preprovisioned processing capacity. Sufficient queue retention preserves affected updates until the partner recovers and the backlog can be processed.

### Question 3 (DVA-C02)

**Question:**

A Python AWS Lambda function handles synchronous requests for both shipment tracking and shipping-label creation. The handler routes requests to the appropriate third-party API. The function has reserved concurrency of 100. Tracking requests require no more than 20 concurrent executions at peak load.

When the label provider becomes slow, label requests occupy all 100 concurrent executions, and tracking requests are throttled even though the tracking provider is healthy. The two request types share no other constrained resources. The application must support peak tracking traffic during a label-provider slowdown while limiting the combined concurrency of these workloads to 100.

Which change should the developer make?

**Correct answer:** Deploy tracking and label creation as separate Lambda functions and route requests accordingly. Set reserved concurrency to 20 for tracking and 80 for label creation.

**Explanation:** Separate functions with independent reserved concurrency allocations create capacity boundaries between the workloads. Label creation cannot consume tracking's 20 reserved executions, and the two function limits keep combined concurrency at or below 100.

### Question 4 (DOP-C02)

**Question:**

A SaaS company runs a transaction-processing application on Amazon EC2 behind an Application Load Balancer. One Auto Scaling group spans three Availability Zones, and all tenants share an Amazon RDS for PostgreSQL Multi-AZ DB instance. During an incident, expensive requests from one tenant exhausted the application's worker capacity and the database's connections, interrupting service for every tenant.

The company can divide tenants into 10 fixed groups with approximately equal workloads. It requires resource exhaustion caused by one tenant to affect only that tenant's group. Every group must also remain able to process transactions after any single Availability Zone fails, using existing surviving compute capacity. Brief interruptions during automatic database failover are acceptable. Moving tenants between groups during an incident is not permitted.

Which architecture meets these requirements?

**Correct answer:** Create a separate application deployment for each tenant group, with its own Auto Scaling group spanning three Availability Zones and its own RDS Multi-AZ DB instance. Route each tenant only to its assigned deployment, eliminate shared runtime dependencies between deployments, and provision each deployment with enough surviving compute capacity for an Availability Zone failure.

**Explanation:** Each tenant group has independent compute and database resources, so exhaustion within one deployment does not consume another deployment's application or database capacity. Distributing each group's compute across Availability Zones with sufficient surviving capacity, together with its own Multi-AZ database, also supports continued operation after a single Availability Zone failure.

### Question 5 (SAA-C03)

**Question:**

A company runs production Amazon EC2 instances in private subnets across two Availability Zones in one AWS Region. Administrators connect through a single bastion host by using SSH. The bastion host's security group accepts SSH only from the company's fixed public IP range, and production security groups accept SSH only from the bastion host's security group.

The company requires administrators to retain SSH access to surviving production instances if either Availability Zone becomes unavailable. Access must not depend on launching replacement instances or changing security group rules during an outage. Production instances must remain inaccessible directly from the internet.

Which design meets these requirements?

**Correct answer:** Run one bastion host in a public subnet in each Availability Zone, each with its own public address. Associate both hosts with the existing bastion security group, and provide administrators with both addresses.

**Explanation:** Each Availability Zone has an already-running administrative access path. If one zone fails, administrators can use the surviving bastion to reach surviving production instances. The shared security group preserves the corporate-source restriction and authorizes both bastions to access production without exposing production instances directly to the internet.

## Practice Fault Isolation in a full exam

Questions on Fault Isolation 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): 8 questions
- [SOA-C03](https://blowstack.com/practice-exams/aws-certified-cloudops-engineer-associate): 5 questions
- [SAP-C03](https://blowstack.com/practice-exams/aws-certified-solutions-architect-professional): 5 questions
- [CLF-C02](https://blowstack.com/practice-exams/aws-certified-cloud-practitioner): 1 question
- [DVA-C02](https://blowstack.com/practice-exams/aws-certified-developer-associate): 1 question
- [DOP-C02](https://blowstack.com/practice-exams/aws-certified-devops-engineer-professional): 1 question
- [ANS-C01](https://blowstack.com/practice-exams/aws-certified-advanced-networking-specialty): 1 question

## What is Fault Isolation

Fault isolation limits how far a failure can spread. A failure domain is a set of resources that can be affected by the same event, such as a host failure, power interruption, or faulty deployment. Good architecture identifies these domains and separates resources so that one localized failure does not disable every component needed to serve a request. The scope of the resulting impact is often called the blast radius.

Cloud infrastructure provides several isolation boundaries, including physical hosts, Availability Zones, and Regions. In AWS, an Availability Zone consists of one or more discrete data centers with redundant power, networking, and connectivity. Availability Zones within a Region are physically separated and designed not to share single points of failure in their supporting infrastructure. Consequently, a localized infrastructure failure affecting one zone should not also disable another zone through the same failed infrastructure component. Low-latency links between zones support applications that distribute compute and replicate data across these boundaries.

For example, consider an application with web servers in two Availability Zones behind a load balancer. If one zone becomes unavailable, the load balancer can direct requests to healthy servers in the other zone. This only preserves the application if its other dependencies also remain usable. A database, shared file system, outbound network path, or authentication dependency confined to the failed zone can still interrupt the entire workload. Engineers must examine the complete request path rather than count only the number of web servers.

Isolation and redundancy solve related but different problems. Isolation contains a failure; redundancy supplies alternatives. Two virtual machines on the same physical host provide less protection from host failure than machines on separate hosts. Similarly, multiple servers in one zone do not protect against loss of that zone. Deploying across zones does not automatically configure database replication, traffic failover, or adequate surviving capacity. These require explicit architectural choices or appropriately configured managed services.

Fault isolation also applies to software and operations. Separate worker pools can prevent one customer's expensive jobs from exhausting every worker. Timeouts and circuit breakers can stop a failed dependency from consuming all application connections. Staggered deployments can keep a defective release from reaching every zone simultaneously. Infrastructure separation cannot contain a configuration error that is deliberately applied everywhere at once.

The trade-offs include additional infrastructure cost, cross-zone transfer charges where applicable, replication latency, and greater operational complexity. Surviving resources need enough capacity to absorb redirected traffic, and database failover may interrupt connections even when data remains protected. Test zone-loss scenarios, connection recovery, and dependency behavior rather than assuming that a multi-zone diagram proves resilience. For certification questions, recognize Availability Zones as isolated infrastructure failure domains within a Region, not as individual servers or synonyms for Regions. Their separation reduces shared failure exposure; it does not guarantee that every application failure or regional event is isolated.

## Learn Fault Isolation in lessons

7 short lessons on Fault Isolation in 5 certification courses.

- CLF-C02: Fault Isolation: Availability Zone failure domains (Premium), https://app.blowstack.com/lessons/fault-isolation-availability-zone-failure-domains-clf-c02
- SOA-C03: Fault Isolation and Fault Tolerance: containment versus continued operation (Premium), https://app.blowstack.com/lessons/fault-isolation-and-fault-tolerance-containment-versus-continued-operation-soa-c03
- SAP-C03: Fault Isolation: independent cluster failure domains (Premium), https://app.blowstack.com/lessons/fault-isolation-independent-cluster-failure-domains-sap-c03
- DOP-C02: Fault Isolation vs High Availability (Premium), https://app.blowstack.com/lessons/fault-isolation-vs-high-availability-dop-c02
- SAA-C03: Fault Isolation: AWS infrastructure failure domains (Premium), https://app.blowstack.com/lessons/fault-isolation-aws-infrastructure-failure-domains-saa-c03
- SOA-C03: Fault Isolation: Availability Zone failure domains (Premium), https://app.blowstack.com/lessons/fault-isolation-availability-zone-failure-domains-soa-c03
