Skip to content

Service Comparison - Serverless

Overview

This guide compares serverless compute, event routing, workflow orchestration, and API management services across AWS, Azure, and Google Cloud. Serverless architecture is a key topic across all cloud certification exams.

Serverless Compute

Lambda vs Azure Functions vs Cloud Functions

Feature AWS Lambda Azure Functions Google Cloud Functions
Runtimes Node.js, Python, Java, .NET, Go, Ruby, custom Node.js, Python, Java, .NET, PowerShell, custom Node.js, Python, Java, .NET, Go, Ruby, PHP
Max Execution Time 15 minutes 10 minutes (Consumption), unlimited (Premium/Dedicated) 9 minutes (1st gen), 60 minutes (2nd gen)
Max Memory 10,240 MB 1,536 MB (Consumption), 14 GB (Premium) 32 GB (2nd gen)
Max Package Size 250 MB (unzipped), 50 MB (zipped) No hard limit (Consumption: plan limits) 500 MB (source), 32 GB (container)
Container Support Container images (up to 10 GB) Custom containers (Premium/Dedicated) Container images (2nd gen)
Ephemeral Storage 512 MB - 10 GB (/tmp) 500 MB - 1 TB (local storage) In-memory tmpfs
Concurrency Model 1 invocation per instance (default) Multiple invocations per instance 1 invocation per instance (1st gen), concurrent (2nd gen)
Max Concurrency 1,000 (default, adjustable) 200 per instance (Consumption) 1,000 (per function, 2nd gen)
Reserved Concurrency Yes (provisioned concurrency) Pre-warmed instances (Premium) Min instances (2nd gen)
Cold Start ~100ms-1s (varies by runtime) ~1-3s (Consumption), reduced (Premium) ~100ms-2s (1st gen), reduced (2nd gen)
VPC Access VPC configuration (ENI-based) VNet integration VPC connector / Direct VPC
GPU Support No No No
Pricing Per request + per GB-second Per execution + per GB-second Per invocation + per GB-second + per GHz-second
Free Tier 1M requests + 400,000 GB-seconds/month 1M executions + 400,000 GB-seconds/month 2M invocations + 400,000 GB-seconds/month
SnapStart Yes (Java) N/A N/A

Cold Start Optimization Strategies

Strategy AWS Lambda Azure Functions Google Cloud Functions
Provisioned/Pre-warmed Provisioned Concurrency Premium Plan (pre-warmed) Min instances (2nd gen)
Snapshots SnapStart (Java) N/A N/A
Keep-Warm Scheduled ping Durable Functions timer Cloud Scheduler trigger
Container Reuse Execution context reuse Instance reuse Instance reuse
Smaller Packages Layer optimization Deployment optimization Dependency optimization
Runtime Choice Python/Node.js (fastest) Node.js/.NET (fastest) Node.js/Python (fastest)

Serverless Compute - Extended Options

Lambda vs Azure Container Apps vs Cloud Run

Feature AWS Lambda Azure Container Apps Google Cloud Run
Unit of Deployment Function (code/container) Container Container
Scale to Zero Yes Yes Yes
Max Request Duration 15 minutes Unlimited (HTTP), 24h (jobs) 60 minutes
Max vCPU 6 vCPU 4 vCPU 8 vCPU
Max Memory 10 GB 8 GB 32 GB
Concurrency per Instance 1 (default) Multiple Up to 1,000
GPU Support No No Yes (L4)
Built-in Auth IAM + Cognito (via API Gateway) Easy Auth IAM
Traffic Splitting Aliases + weighted Revision-based Revision-based
Sidecar Support Lambda Extensions Yes Yes
gRPC No Yes Yes
WebSocket Via API Gateway Yes No (use Firebase)

Event Routing and Integration

EventBridge vs Event Grid vs Eventarc

Feature AWS EventBridge Azure Event Grid Google Eventarc
Type Serverless event bus Event routing service Event routing service
Event Sources 200+ AWS services, SaaS, custom Azure services, custom topics, partner 130+ Google Cloud sources, custom
Event Format CloudEvents + custom CloudEvents (v1.0) CloudEvents (v1.0)
Targets 20+ (Lambda, SQS, SNS, Step Functions, API destinations) Azure Functions, Logic Apps, webhooks, Event Hubs, Storage Cloud Run, Cloud Functions, Workflows, GKE
Filtering Content-based (prefix, suffix, numeric, IP, exists) Subject, event type, advanced (AND/OR) Attribute matching
Schema Registry Yes (discovered + custom) Yes (Event Grid schema) N/A
Event Replay Archive and replay N/A N/A
Dead Letter Queue Yes (SQS) Yes (Blob Storage) Yes (Pub/Sub)
Cross-Account Cross-account event buses Cross-subscription topics Cross-project
Global Regional (multi-region with global endpoints) Regional Regional
Pricing Per event published Per operation Per event delivered
Free Tier All AWS service events free First 100K operations/month First 2M events/month
Guaranteed Delivery At-least-once At-least-once At-least-once
Latency ~500ms (typical) ~seconds ~seconds

Event Bus vs Message Queue Decision

Use Case Event Routing (EventBridge/Event Grid/Eventarc) Message Queue (SQS/Service Bus/Pub/Sub)
Fan-out Yes (multiple targets per rule) Pub/Sub pattern
Point-to-point Not ideal Yes
Ordering Best-effort FIFO available
Retention Limited (archive for replay) Days to weeks
Replay EventBridge archive/replay Dead letter reprocessing
Throughput High (millions of events) Very high (millions of messages)
Transformation Input transformers Consumer-side

Workflow Orchestration

Step Functions vs Logic Apps/Durable Functions vs Workflows

Feature AWS Step Functions Azure Logic Apps Google Cloud Workflows
Type State machine workflow Visual workflow designer YAML/JSON workflow
Definition Amazon States Language (JSON) Logic Apps Designer / Bicep Workflows YAML syntax
Max Execution 1 year (Standard), 5 min (Express) 90 days (Consumption) 1 year
Execution Model Standard (exactly-once), Express (at-least-once) Stateful, Stateless Exactly-once
Visual Designer Workflow Studio Logic Apps Designer (visual) N/A (YAML only)
Connectors AWS service integrations (200+) 1,000+ connectors (SaaS) HTTP, gRPC, Cloud APIs
Human Approval Callback pattern (task tokens) Approval connector Callback endpoints
Parallel Execution Parallel state, Map state Parallel branches, ForEach Parallel steps
Error Handling Retry, Catch Retry policies, scopes Retry, try/except
Nested Workflows Yes Yes Yes (subworkflows)
Express/Sync Express Workflows Stateless N/A
Pricing Per state transition (Standard), per execution + duration (Express) Per action execution, per trigger Per step executed
Free Tier 4,000 transitions/month 4,000 actions/month (Consumption) 5,000 steps/month

Durable Functions vs Step Functions

Feature Azure Durable Functions AWS Step Functions
Language Code-based (C#, JavaScript, Python, Java, PowerShell) JSON (Amazon States Language)
Patterns Orchestrator, Entity, Client State machine
Fan-out/Fan-in Native pattern Map state
Human Interaction WaitForExternalEvent Task tokens (callback)
Eternal Orchestrations ContinueAsNew Loop with choice state
Entity Pattern Durable Entities N/A (use DynamoDB)
Sub-Orchestrations Yes Nested execution
Testing Unit testable Limited (local testing tool)

API Management

API Gateway vs Azure API Management vs Apigee

Feature AWS API Gateway Azure API Management Google Apigee
Type Managed API gateway Full API lifecycle management Full API lifecycle management
Protocols REST, WebSocket, HTTP REST, SOAP, WebSocket, GraphQL, gRPC REST, SOAP, gRPC, GraphQL
Developer Portal N/A (use third-party) Built-in Built-in
API Products N/A Products (bundled APIs) API Products
Rate Limiting Usage plans + API keys Policies (rate-limit, quota) Spike arrest, quota
Caching Built-in (0.5-237 GB) Built-in, External (Redis) Built-in, External
Request Transform Mapping templates (VTL) Policies (XML-based) Policies (JavaScript, Python)
Response Transform Mapping templates Policies Policies
Authentication IAM, Cognito, Lambda authorizer OAuth 2.0, API keys, certificates, JWT OAuth 2.0, API keys, JWT, SAML
mTLS Yes Yes Yes
WAF Integration AWS WAF N/A (use Azure WAF separately) Built-in threat protection
Monetization N/A N/A (third-party) Built-in monetization
Analytics CloudWatch Built-in analytics Built-in analytics (Apigee Analytics)
Pricing Model Per request + data transfer Tier-based (Developer to Premium) Per API call (Pay-as-you-go) or subscription

Serverless Storage

Serverless Database Options

Feature DynamoDB Cosmos DB (Serverless) Firestore
Type Key-value + document Multi-model Document
Consistency Eventually / Strong (per-item) 5 consistency levels Strong
Auto-Scaling On-demand / provisioned Serverless (RU-based) Automatic
Global Distribution Global Tables Multi-region writes Multi-region (active-passive)
Transactions ACID (up to 100 items) ACID (within partition) ACID (up to 500 docs)
Max Item Size 400 KB 2 MB 1 MB
Secondary Indexes GSI, LSI Automatic indexing Composite indexes
Change Stream DynamoDB Streams Change Feed Real-time listeners
Pricing Per RCU/WCU or per request Per RU consumed + storage Per read/write/delete + storage
Free Tier 25 GB + 25 RCU/WCU 1,000 RU/s + 25 GB 1 GB + 50K reads + 20K writes/day

Serverless Integration Patterns

Event-Driven Architecture Components

Pattern AWS Azure GCP
Event Bus EventBridge Event Grid Eventarc
Message Queue SQS Service Bus Queue Pub/Sub
Stream Processing Kinesis / MSK Serverless Event Hubs Pub/Sub + Dataflow
Pub/Sub SNS Service Bus Topic Pub/Sub
Scheduling EventBridge Scheduler Timer trigger (Functions) Cloud Scheduler
API Gateway API Gateway API Management Apigee / API Gateway
Workflow Step Functions Logic Apps / Durable Functions Workflows
File Processing S3 + Lambda Blob Storage + Functions Cloud Storage + Cloud Functions
WebSocket API Gateway WebSocket SignalR Service Firebase Realtime DB
GraphQL AppSync N/A (use Hot Chocolate / Strawberry Shake) N/A (use Apollo on Cloud Run)

Certification Exam Focus Areas

AWS Solutions Architect / Developer

  • Lambda execution model, concurrency, and cold starts
  • EventBridge event patterns and rule matching
  • Step Functions Standard vs Express workflows
  • API Gateway stage variables, deployment, and caching
  • DynamoDB capacity modes and partition key design
  • Lambda layers, extensions, and SnapStart

Azure Developer / Solutions Architect

  • Azure Functions hosting plans (Consumption, Premium, Dedicated)
  • Durable Functions orchestration patterns
  • Event Grid event subscriptions and filtering
  • Logic Apps connectors and workflow design
  • Cosmos DB serverless vs provisioned throughput
  • API Management policies and products

Google Cloud Architect / Developer

  • Cloud Functions 1st gen vs 2nd gen differences
  • Cloud Run concurrency and autoscaling configuration
  • Eventarc triggers and event routing
  • Workflows syntax and error handling
  • Firestore data model and security rules
  • Apigee API proxy design
  • AWS Lambda: https://docs.aws.amazon.com/lambda/latest/dg/
  • AWS Step Functions: https://docs.aws.amazon.com/step-functions/latest/dg/
  • AWS EventBridge: https://docs.aws.amazon.com/eventbridge/latest/userguide/
  • AWS API Gateway: https://docs.aws.amazon.com/apigateway/latest/developerguide/
  • Azure Functions: https://learn.microsoft.com/en-us/azure/azure-functions/
  • Azure Logic Apps: https://learn.microsoft.com/en-us/azure/logic-apps/
  • Azure Event Grid: https://learn.microsoft.com/en-us/azure/event-grid/
  • Google Cloud Functions: https://cloud.google.com/functions/docs
  • Google Cloud Run: https://cloud.google.com/run/docs
  • Google Eventarc: https://cloud.google.com/eventarc/docs
  • Google Cloud Workflows: https://cloud.google.com/workflows/docs

Key Takeaways

  1. Cloud Run and Azure Container Apps blur the line between serverless and containers - offering scale-to-zero with container flexibility
  2. Lambda has the most mature ecosystem but is constrained to 15-minute execution time
  3. Azure Functions Premium Plan eliminates cold starts - unique among the big three
  4. EventBridge leads in event routing with schema registry and archive/replay capabilities
  5. Logic Apps offers the most pre-built connectors (1,000+) for enterprise integration
  6. Step Functions provides the cleanest workflow-as-state-machine model
  7. For exam prep, understand the trade-offs between event-driven and request-driven architectures
  8. All three providers are converging on CloudEvents as the standard event format

Related Guides: - Compute Service Comparison - Containers and Kubernetes Comparison - Messaging and Queues Comparison - Database Service Comparison - Topic: Serverless