AWS Service Comparisons¶
Compute Services¶
| Service | Use Case | Pros | Cons | When to Choose |
|---|---|---|---|---|
| EC2 | Virtual servers | Full control, wide instance types | Manage OS, patching | Need custom configurations |
| Lambda | Event-driven functions | Serverless, automatic scaling | 15min limit, cold starts | Short, stateless workloads |
| ECS | Container orchestration | AWS-native, Fargate option | Learning curve | Containerized applications |
| EKS | Kubernetes | Industry standard | Complex, expensive | Kubernetes expertise |
| Batch | Batch job processing | Managed queuing, spot support | Not real-time | Large-scale batch jobs |
| Lightsail | Simple VPS | Easy setup, predictable pricing | Limited features | Simple web applications |
Storage Services¶
| Service | Type | Use Case | Durability | Access Pattern |
|---|---|---|---|---|
| S3 Standard | Object | Frequently accessed data | 99.999999999% | Regular access |
| S3 IA | Object | Infrequently accessed | 99.999999999% | Monthly access |
| S3 One Zone-IA | Object | Reproducible, infrequent | 99.999999999% | Monthly, no AZ resilience |
| S3 Glacier | Object | Archive, backup | 99.999999999% | Minutes to hours retrieval |
| S3 Deep Archive | Object | Long-term archive | 99.999999999% | 12+ hours retrieval |
| EBS | Block | EC2 instance storage | 99.999% | High IOPS, low latency |
| EFS | File | Shared file storage | 99.999999999% | Concurrent access |
| FSx | File | High-performance workloads | 99.999999999% | Windows, Lustre |
Database Services¶
| Service | Type | Engine | Use Case | Scaling |
|---|---|---|---|---|
| RDS | Relational | MySQL, PostgreSQL, etc. | OLTP, complex queries | Vertical, read replicas |
| Aurora | Relational | MySQL, PostgreSQL compatible | High performance OLTP | Auto-scaling |
| DynamoDB | NoSQL | Proprietary | Single-digit ms latency | Horizontal |
| DocumentDB | NoSQL | MongoDB compatible | Document storage | Horizontal |
| Redshift | Data Warehouse | PostgreSQL compatible | OLAP, analytics | Vertical, concurrency scaling |
| ElastiCache | In-memory | Redis, Memcached | Caching, session store | Horizontal |
| Neptune | Graph | Gremlin, SPARQL | Graph relationships | Vertical |
| Timestream | Time-series | Proprietary | IoT, metrics | Auto-scaling |
Networking Services¶
| Service | Purpose | Layer | Use Case |
|---|---|---|---|
| ALB | Load balancing | Layer 7 (HTTP/HTTPS) | Web applications, microservices |
| NLB | Load balancing | Layer 4 (TCP/UDP) | High performance, static IP |
| CLB | Load balancing | Layer 4/7 | Legacy applications |
| CloudFront | CDN | Layer 7 | Global content delivery |
| Route 53 | DNS | Layer 7 | Domain resolution, health checks |
| API Gateway | API management | Layer 7 | REST/WebSocket APIs |
| Direct Connect | Network connection | Layer ½ | Dedicated network link |
| VPN | Network connection | Layer 3 | Encrypted connection over internet |
Security Services¶
| Service | Purpose | Scope | Key Features |
|---|---|---|---|
| IAM | Identity & Access | Global | Users, roles, policies |
| Cognito | User authentication | Regional | User pools, identity pools |
| KMS | Key management | Regional | Encryption key management |
| CloudHSM | Hardware security | AZ | Dedicated HSM instances |
| WAF | Web application firewall | Global/Regional | SQL injection, XSS protection |
| Shield | DDoS protection | Global | Standard (free), Advanced (paid) |
| GuardDuty | Threat detection | Regional | ML-based anomaly detection |
| Inspector | Security assessment | Regional | Application security |
Monitoring & Management¶
| Service | Purpose | Scope | Key Metrics |
|---|---|---|---|
| CloudWatch | Monitoring | Regional | Metrics, logs, alarms |
| X-Ray | Application tracing | Regional | Request tracing, performance |
| CloudTrail | API logging | Global/Regional | Governance, compliance |
| Config | Configuration compliance | Regional | Resource compliance |
| Systems Manager | System management | Regional | Patch management, automation |
| Trusted Advisor | Best practices | Global | Cost, security, performance |
Integration Services¶
| Service | Pattern | Message Type | Durability |
|---|---|---|---|
| SQS | Queue | Point-to-point | Persistent |
| SNS | Pub/Sub | Fanout | Transient |
| EventBridge | Event bus | Event-driven | Persistent |
| Step Functions | Workflow | State machine | Persistent |
| SWF | Workflow | Task coordination | Persistent |
| MQ | Message broker | Enterprise messaging | Persistent |
| AppSync | GraphQL | Real-time data | Persistent |
| Kinesis | Streaming | Real-time data streams | Configurable retention |
Cost Comparison Guidelines¶
Compute Cost Factors¶
- On-Demand: Highest cost, maximum flexibility
- Reserved Instances: 40-75% savings, 1-3 year commitment
- Spot Instances: Up to 90% savings, can be interrupted
- Savings Plans: Flexible discount, compute usage commitment
Storage Cost Factors¶
- Frequent Access: S3 Standard > S3 IA > S3 One Zone-IA
- Archive: Glacier > Deep Archive (cheaper storage, higher retrieval cost)
- Performance: Higher IOPS = higher cost
- Durability: Higher durability = higher cost
Database Cost Factors¶
- Serverless: DynamoDB On-Demand > Aurora Serverless > Traditional
- Reserved Capacity: DynamoDB Reserved > RDS Reserved
- Read Scaling: Read replicas add cost but improve performance
- Multi-AZ: Doubles storage and compute costs
Service Selection Decision Tree¶
For Compute:¶
- Need full OS control? → EC2
- Event-driven, < 15 minutes? → Lambda
- Containerized application? → ECS/EKS
- Batch processing? → AWS Batch
- Simple web application? → Lightsail
For Storage:¶
- Object storage? → S3 (choose class by access pattern)
- Block storage for EC2? → EBS
- Shared file system? → EFS
- High-performance file system? → FSx
For Database:¶
- Complex relational queries? → RDS/Aurora
- Simple, fast NoSQL? → DynamoDB
- Analytics/Data warehouse? → Redshift
- In-memory caching? → ElastiCache
- Graph database? → Neptune
For Load Balancing:¶
- HTTP/HTTPS with advanced routing? → ALB
- High performance, static IP? → NLB
- Simple TCP/UDP? → CLB or NLB
- Legacy application? → CLB