Cloud Patterns

Overview

Moving applications to the cloud does not automatically make them scalable, resilient, secure, or cloud-native.

Many organizations discover that the same architectural problems still exist after migration.

Cloud architects are frequently asked questions such as:

Cloud patterns provide proven solutions to these recurring challenges.

A Running Example

Throughout this page we will use a healthcare diagnostics platform undergoing cloud modernization.

The organization starts with a traditional monolithic application running on-premises.

Every cloud pattern introduced throughout this page supports this modernization journey.

Why Cloud Patterns Matter

Cloud infrastructure provides powerful capabilities, but architecture decisions still determine success or failure.

Cloud patterns help organizations build systems that scale, recover, evolve, and operate more effectively.

Cloud Challenges

Cloud patterns exist because organizations repeatedly encounter similar cloud architecture challenges.

Legacy Modernization

Most enterprises cannot rewrite decades of software overnight.

Traffic Spikes

Demand may increase dramatically without warning.

Dependency Failures

Cloud services, APIs, and external systems eventually fail.

Operational Complexity

As cloud solutions grow, operational management becomes increasingly difficult.

Security Boundaries

Resources must be protected while remaining accessible.

Cost Optimization

Cloud resources can scale rapidly, and so can cloud bills.

Understanding these challenges is often more important than memorizing pattern names.

Choosing A Cloud Pattern

Cloud patterns should be selected based on business objectives and architectural constraints.

The goal is not to implement cloud patterns everywhere. The goal is to solve cloud architecture problems effectively.

Strangler Fig Pattern

Few organizations can afford a complete system rewrite.

The Strangler Fig Pattern enables gradual modernization by replacing parts of a legacy application incrementally.

Instead of a risky big-bang migration, functionality is migrated one capability at a time.

Benefits:

• Reduced Modernization Risk
• Incremental Delivery
• Faster Business Value
• Lower Migration Complexity

Challenges:

• Temporary Hybrid Architecture
• Long Transition Periods
• Integration Complexity

Works Well When:

• Large Legacy Systems Exist
• Incremental Migration Is Preferred
• Business Cannot Tolerate Large Rewrites

Avoid When:

• System Is Small Enough To Replace Directly
• Legacy System Can Be Retired Quickly

Sidecar Pattern

The Sidecar Pattern moves supporting capabilities outside application code while keeping them close to the application.

Common responsibilities include monitoring, logging, security, service discovery, and telemetry.

Benefits:

• Separation Of Concerns
• Operational Consistency
• Reduced Application Complexity
• Reusable Platform Services

Challenges:

• Additional Resource Usage
• Deployment Complexity
• Operational Overhead

Works Well When:

• Cross-Cutting Concerns Exist
• Kubernetes Platforms Are Used
• Platform Standardization Is Important

Avoid When:

• Applications Are Extremely Simple
• Additional Components Add Little Value

Ambassador Pattern

Applications often communicate with external services, APIs, and cloud resources.

The Ambassador Pattern introduces a proxy component that handles external communication on behalf of the application.

The application interacts with the ambassador while the ambassador manages network concerns.

Benefits:

• Simplified Application Logic
• Centralized Communication Policies
• Improved Resilience
• Reusable Connectivity Logic

Challenges:

• Additional Infrastructure
• Operational Management
• Extra Network Hop

Works Well When:

• Many External Dependencies Exist
• Connection Policies Need Standardization
• Network Complexity Must Be Hidden

Avoid When:

• External Integrations Are Minimal
• Network Complexity Does Not Exist

Anti-Corruption Layer

Modern cloud-native systems often need to integrate with legacy platforms.

The Anti-Corruption Layer prevents legacy concepts and data models from contaminating modern architectures.

The layer translates requests, responses, and business concepts between systems.

Benefits:

• Protects Modern Architecture
• Isolates Legacy Complexity
• Supports Gradual Modernization
• Improves Maintainability

Challenges:

• Additional Components
• Mapping Complexity
• Maintenance Requirements

Works Well When:

• Legacy Integration Exists
• Domain Models Differ Significantly
• Modernization Is Ongoing

Avoid When:

• Systems Already Share Similar Models
• Integration Is Short Lived

Gatekeeper Pattern

Organizations often need a controlled entry point into cloud resources.

The Gatekeeper Pattern ensures requests are validated before reaching protected services.

This enables enforcement of security, compliance, validation, and governance requirements.

Benefits:

• Improved Security
• Centralized Policy Enforcement
• Reduced Exposure
• Better Governance

Challenges:

• Additional Latency
• Operational Overhead
• Potential Bottlenecks

Works Well When:

• Sensitive Resources Exist
• Strict Access Controls Are Needed
• Compliance Requirements Exist

Avoid When:

• Resources Are Already Isolated Effectively
• Additional Controls Provide Limited Value

Gateway Routing Pattern

As applications grow, exposing every service directly becomes difficult to manage.

The Gateway Routing Pattern introduces a centralized entry point that routes requests to appropriate services.

The gateway becomes the controlled front door to the platform.

Benefits:

• Simplified Client Access
• Centralized Security
• Centralized Routing
• Improved Governance

Challenges:

• Additional Infrastructure
• Gateway Scaling Requirements
• Potential Single Entry Dependency

Works Well When:

• Many Services Exist
• Multiple Clients Exist
• Centralized Policies Are Needed

Avoid When:

• Only A Few Services Exist
• Direct Communication Is Sufficient

Backends For Frontends (BFF)

Different client applications often have different requirements.

Mobile applications, web applications, partner integrations, and internal portals rarely need identical APIs.

The BFF Pattern creates dedicated backends tailored to specific client experiences.

Each backend optimizes data, responses, and workflows for its consumers.

Benefits:

• Optimized User Experience
• Reduced Client Complexity
• Independent Evolution
• Improved Performance

Challenges:

• Additional Services
• Potential Code Duplication
• Operational Complexity

Works Well When:

• Multiple Client Types Exist
• Client Requirements Differ Significantly
• Performance Optimization Matters

Avoid When:

• Clients Have Nearly Identical Requirements
• A Single API Meets All Needs

Queue-Based Load Leveling

Cloud workloads are rarely distributed evenly throughout the day.

Traffic spikes can overwhelm downstream services if every request is processed immediately.

Queue-Based Load Leveling introduces a queue to absorb demand and smooth processing.

The queue acts as a buffer between fast producers and slower consumers.

Benefits:

• Spike Protection
• Improved Stability
• Scalable Processing
• Reduced Dependency Pressure

Challenges:

• Increased Latency
• Queue Monitoring Required
• Backlog Management

Works Well When:

• Traffic Fluctuates Significantly
• Processing Is Resource Intensive
• Asynchronous Processing Is Acceptable

Avoid When:

• Immediate Responses Are Mandatory
• Processing Cannot Be Delayed

Competing Consumers Pattern

Eventually a single consumer becomes a bottleneck.

The Competing Consumers Pattern increases processing capacity by allowing multiple workers to consume messages from the same queue.

Each message is processed by only one consumer.

Benefits:

• Horizontal Scalability
• Higher Throughput
• Improved Resource Utilization
• Faster Work Completion

Challenges:

• Ordering Considerations
• Concurrency Management
• Duplicate Protection Needs

Works Well When:

• Large Message Volumes Exist
• Work Can Be Parallelized
• Scalability Is Required

Avoid When:

• Strict Global Ordering Is Required
• Workloads Are Highly Dependent

Health Endpoint Monitoring

Cloud systems fail in many different ways.

Health Endpoint Monitoring exposes service health information that platforms can use to make operational decisions.

Health checks help determine whether services should receive traffic or be removed from rotation.

Benefits:

• Improved Reliability
• Faster Failure Detection
• Automated Recovery Support
• Better Observability

Challenges:

• Meaningful Health Checks Required
• Monitoring Configuration
• False Positives

Works Well When:

• Service Availability Matters
• Cloud Automation Exists
• Operational Visibility Is Important

Avoid When:

• Endpoints Provide No Useful Information
• Health Checks Are Misleading

Throttling Pattern

Unlimited access to cloud resources can result in instability, outages, and unexpected costs.

The Throttling Pattern controls how much work consumers can perform within a given period.

The objective is protecting systems before they become overloaded.

Benefits:

• Resource Protection
• Improved Stability
• Cost Control
• Fair Resource Usage

Challenges:

• Policy Tuning
• User Experience Considerations
• Capacity Planning

Works Well When:

• Shared Platforms Exist
• External Consumers Exist
• Traffic Growth Is Unpredictable

Avoid When:

• Traffic Volumes Are Minimal
• Resource Constraints Do Not Exist

Bulkhead Pattern

Cloud environments contain many workloads competing for resources.

The Bulkhead Pattern isolates workloads so failures in one area do not impact other critical business functions.

If one workload exhausts its resources, others remain available.

Benefits:

• Failure Isolation
• Improved Availability
• Reduced Blast Radius
• Resource Protection

Challenges:

• Capacity Planning
• Resource Fragmentation
• Operational Complexity

Works Well When:

• Multiple Workloads Share Infrastructure
• Critical Services Require Protection
• Failure Isolation Matters

Avoid When:

• Workloads Are Very Small
• Isolation Adds Excessive Complexity

Circuit Breaker Pattern

Repeatedly calling a failing dependency often makes outages worse.

The Circuit Breaker Pattern stops requests when failures exceed predefined thresholds.

Instead of waiting for repeated failures, systems fail fast and recover more safely.

Benefits:

• Dependency Protection
• Reduced Cascading Failures
• Improved Stability
• Faster Recovery

Challenges:

• Configuration Tuning
• Recovery Management
• Additional Complexity

Works Well When:

• External Dependencies Exist
• Service Failures Are Common
• Availability Matters

Avoid When:

• Dependencies Are Extremely Reliable
• Additional Protection Provides Little Value

Cache-Aside Pattern

Accessing databases for every request increases latency, cost, and infrastructure pressure.

The Cache-Aside Pattern stores frequently requested data in a cache and accesses the database only when necessary.

This pattern is one of the most commonly used cloud optimization techniques.

Benefits:

• Reduced Database Load
• Improved Response Times
• Lower Cloud Costs
• Better Scalability

Challenges:

• Cache Invalidation
• Stale Data
• Consistency Tradeoffs

Works Well When:

• Read Traffic Is High
• Data Changes Less Frequently
• Low Latency Matters

Avoid When:

• Data Changes Constantly
• Strong Consistency Is Required

Valet Key Pattern

Applications often need to provide temporary access to cloud resources without exposing permanent credentials.

The Valet Key Pattern grants limited, time-bound access to specific resources.

This pattern is commonly used for document uploads, downloads, image sharing, and external partner integrations.

Benefits:

• Improved Security
• Reduced Credential Exposure
• Temporary Access Control
• Lower Application Load

Challenges:

• Token Management
• Expiration Handling
• Access Governance

Works Well When:

• External Resource Access Exists
• Large File Transfers Occur
• Temporary Permissions Are Needed

Avoid When:

• Permanent Access Is Required
• Resources Cannot Be Safely Scoped

Cloud Pattern Combinations

Cloud-native platforms rarely rely on a single pattern.

Most successful cloud architectures combine multiple patterns that address different business and operational challenges.

Strangler Fig + Anti-Corruption Layer

This combination enables safe and controlled modernization.

Gateway Routing + BFF

This approach simplifies client experiences while maintaining centralized governance.

Queue-Based Load Leveling + Competing Consumers

This combination supports both workload smoothing and horizontal scalability.

Circuit Breaker + Bulkhead

This combination limits failure propagation and protects critical workloads.

Cache-Aside + Gateway

This combination improves performance and reduces infrastructure pressure.

Pattern Selection Framework

Real-World Modernization Case Study

Consider a healthcare diagnostics platform originally built as a monolithic on-premises application.

Business growth introduces scalability and deployment challenges.

Legacy integration remains necessary during the transition.

Traffic grows significantly.

Dependency resilience becomes critical.

Performance is improved using caching.

Cloud Architecture Review Checklist

Cloud Architecture Canvas

Common Cloud Anti-Patterns

Lift-And-Shift Forever

Applications are moved to the cloud but never modernized.

Cloud-Native In Name Only

Systems run in the cloud but continue operating exactly like traditional on-premises applications.

Scaling Everything

Resources are scaled aggressively without understanding actual bottlenecks.

No Resilience Strategy

Single dependency failures cause major outages.

No Cost Governance

Technically successful cloud adoption becomes financially unsustainable.

Shared Everything

Services share databases, ownership, deployment cycles, and operational responsibilities.

Ignoring Operational Readiness

Monitoring, alerting, and runbooks are treated as afterthoughts.

Designing For Success Only

Systems are optimized for normal operations while failure scenarios are ignored.

How Patterns Connect

Cloud patterns build upon concepts introduced by architectural, integration, messaging, and distributed-system patterns.

• Scalability
• Reliability
• Resilience
• Security
• Performance
• Operational Excellence
• Modernization

As organizations mature their cloud adoption, these patterns become foundational architecture building blocks.

Key Takeaway