AWS — First Experiences with Public Cloud

For fifteen years we ran everything on-premise. Our own servers, our own rack, our own problems. A client with international expansion needed infrastructure in the US and Europe. Time for public cloud. Time for AWS.

EC2: Virtual Servers¶

EC2 instances in minutes instead of weeks (server procurement, delivery, installation). Instance types for every workload — t2.micro for dev, m4.xlarge for production, c4 for compute-intensive workloads. Pay by the hour, scale up and down as needed.

S3: Unlimited Storage¶

Simple Storage Service — object storage with 99.999999999% durability. Backups, static files, archives. Lifecycle policies for automatic transition to cheaper storage classes (Glacier).

VPC: Isolated Network¶

Virtual Private Cloud — your private network in AWS. Subnets (public/private), route tables, security groups, NACLs. Application servers in a private subnet, load balancer in the public one. Similar to on-premise networking, but software-defined.

Lessons from the First Months¶

• Billing alarms: Set them immediately. Forgotten instances = a surprise on the bill
• IAM: Never use the root account. Separate IAM users with least-privilege permissions
• Multi-AZ: Every resource across two Availability Zones. One day it will pay off
• Tags: Tag everything — Environment, Project, Owner. Without tags billing is opaque
• Infrastructure as Code: Terraform/CloudFormation from day one

Cloud is the Future of Infrastructure¶

AWS opened our eyes. Speed of provisioning, global availability, pay-as-you-go. It is not necessarily cheaper than on-premise (it depends on the workload), but it is far more flexible.