Hosting Deep Learning apps on AWS (Stage 3): Infrastructure as Code (IaC) - Serverless, EC2 & Linux

Having successfully deployed a react app to AWS Lambda in Stage 1 and built a local Deep Learning Flask app in Stage 2, we now put these two components together using Serverless as Infrastructure as Code

Here’s the process:

i. Create EC2 instance

In the AWS Console, go to the url below, launch an EC2 (Linux) instance, choosing t3.small rather than t3.micro then create a key pair pem file.

Note that this EC2 machine is not on the free tier (1 GB RAM is definitely not enough!) but it is relatively cheap costing ~ 0.50c per day.

Although t3.small is sufficient for the app, the processor is only 2 GB RAM, so you may want to use a higher grade instance which will of course incur higher charges. For other EC2 instance types see here.

ii. Connect to EC2

After creating your EC2 instance, click on

Instance ID > EC2 Instance Connect tab (default) > Connect

to open a Linux window

iii. Create a project folder on EC2 machine

mkdir lambda-dl-aws

iv. Install Git

Still at the project root folder (i.e. the top level), run the command below to install git

sudo yum install git -y # installs git

v. Install serverless

Also run the command below to install serverless in your lambda-dl-aws folder:

curl -o- -L https://slss.io/install | bash

Check serverless is installed by logging out of your EC2 (Linux) instance and reconnecting, then run:

serverless -version

vi. Install Docker

Besides Git, we need to install Docker on the EC2 machine. Run the following one by one in the aws-lambda-dl-app folder:

sudo yum update -y
sudo yum -y install docker
sudo service docker start
sudo usermod -a -G docker ec2-user
exit

Finally close the Linux window and reconnect to your EC2 instance

Stage 4: Building the Container Image