Cloud Service Models
| Model | আপনি | Provider |
|---|---|---|
| IaaS | OS, app, data | Hardware, virt |
| PaaS | App, data | OS, runtime, scale |
| SaaS | Config, use | সব |
উদাহরণ: IaaS — EC2, VM। PaaS — Heroku, App Engine। SaaS — Gmail, Dropbox।
AWS Core Services ⭐
| Category | Service |
|---|---|
| Compute | EC2, Lambda, ECS, EKS, Fargate |
| Storage | S3, EBS, EFS, Glacier |
| Database | RDS, Aurora, DynamoDB, Redshift, ElastiCache |
| Network | VPC, Route 53, CloudFront, ALB, API Gateway |
| Security | IAM, Cognito, KMS, WAF, Secrets Manager |
| Monitoring | CloudWatch, X-Ray |
| Messaging | SQS, SNS, EventBridge, Kinesis |
| DevOps | CodePipeline, CodeBuild, CloudFormation |
Azure Equivalent
VM, Functions, AKS, Blob Storage, SQL DB, Cosmos DB, VNet, Front Door, Entra ID (AD), Monitor।
GCP Equivalent
Compute Engine, Cloud Functions, GKE, Cloud Storage, Cloud SQL, Spanner, Firestore, BigQuery, Cloud Pub/Sub।
Serverless ⭐
# AWS Lambda
exports.handler = async (event) => {
return {
statusCode: 200,
body: JSON.stringify({ msg: "Hello" })
};
};
# Event sources:
# API Gateway (HTTP), S3 (upload), DynamoDB (stream),
# SQS, SNS, EventBridge (cron), Kinesis
Serverless সুবিধা
- No server management
- Auto-scale 0 → millions
- Pay per execution
- Fast deployment
অসুবিধা
- Cold start (100ms-3s)
- 15min max execution
- Vendor lock-in
- Limited runtime
- Debugging হার্ড
Well-Architected Framework (AWS 6 Pillars)
- Operational Excellence
- Security
- Reliability
- Performance Efficiency
- Cost Optimization
- Sustainability
High Availability Patterns
- Multi-AZ (different data centers)
- Multi-Region (disaster recovery)
- Auto Scaling Group
- Load Balancer
- RDS Multi-AZ replica
- S3 cross-region replication
Infrastructure as Code ⭐
# Terraform
resource "aws_instance" "web" {
ami = "ami-abc123"
instance_type = "t3.micro"
tags = { Name = "web-server" }
}
resource "aws_s3_bucket" "logs" {
bucket = "my-app-logs"
}
Tools: Terraform, Pulumi (code), CloudFormation (AWS), Bicep (Azure), CDK (AWS/Azure)।
FinOps (Cost Optimization) ⭐
- Right-sizing instances
- Reserved Instances (1-3 yr, 40-70% off)
- Spot Instances (90% off, interruptible)
- Savings Plans (flexible)
- Auto-shutdown dev/staging off-hours
- S3 lifecycle → Glacier (archive)
- Delete unused resources (snapshots, EIPs)
- CDN vs direct serve (bandwidth cost)
- Data transfer — cross-region expensive
- Monitor: Cost Explorer, budgets, alerts
- Tag resources by team/project
Multi-Cloud Strategy
| Pros | Cons |
|---|---|
| Avoid lock-in | Complexity |
| Best-of-breed | Skill demand |
| Regulatory | Network cost |
| Negotiation leverage | Ops overhead |
Reality: most use 1 primary + 1 backup (disaster recovery)।
Hybrid Cloud
On-premise + cloud। Regulatory data on-prem, elastic workload cloud-এ। AWS Outposts, Azure Arc।
Shared Responsibility Model
Provider responsible: Of the cloud (hardware, network, hypervisor)
You responsible: In the cloud (OS patching in EC2, data encryption,
IAM, firewall rules, app security)