Scalable App Deployment with AWS
Last updated
Last updated
Let's say we have a web application named MyApp and we want to deploy it to AWS in a scalable fashion. How do we do it?
As described in above diagram the deployment architecture of AWS consist of following components:
MyApp VPC
All EC2 instances are placed inside MyApp VPC for security purposes. MyApp VPC consists of 4 subnets:
Private-1a (Private subnet in AZ- 1a)
Private-1b (Private subnet in AZ- 1b)
Public-1a (Public subnet in AZ- 1a)
Public-1b (Public subnet in AZ- 1b)
Private Subnet - Outer services can not access the instance directly. The instances do not have a public IP. Whereas, the instances themselves can access the internet and any outer services via NAT Internet Gateway.
Public Subnet - Outer Services can access such instances directly through their public IP.
Database Layer
A single node of MongoDB instance will suffice for MyApp.
It can be a Primary Node hosted in Private Subnet of VPC.
A sample setup may look like:
MongoDB - PrimaryNode
Availability Zone: ap-southeast-1a
Instance Size : M3.medium (1 vCPU, 3.75GB RAM, 60GB EBS, Volume attached)
For High Availability (HA) there should be one more DB instance in another Availability Zone (AZ) i.e 1b which can act as the replica for PrimaryNode.
MyApp NodeJS Application Layer
This layer Exists in Private Subnets (1a, 1b) of VPC i.e they are not directly accessible like DB layer.
This is the layer where MyApp is deployed via AWS EC2 Container Service.
A sample Single EC2 instance configuration in this layer:
Instance Size : T2.micro (1 vCPU, 1GB RAM, 20GB EBS)
The instances are kept in an AutoScaling Group where desired Instances are 2 and ECS Service desired Running tasks are 2.
Count is kept at 2 deliberately, for smooth deployment.
LoadBalancer (LB) is hosted in Public Subnets of VPC For High Availability. Ports 80 & 443 of LB are mapped with port 3000 of EC2 instances running MyApp.
CloudFront is placed for caching the static contents like JS and CSS files next to the LB which will serve them and reduce requests to the Original source.
