AWS Classroomnotes 24/Feb/2023

AWS EC2 contd…

Horizontal vs Vertical Scaling

• Differences
  
• Vertical scaling in AWS is possible by changing instance type.
• When we change the instance type the ec2 instance should be in stopped state
• Changing instance type in AWS. Shutdown the ec2 instance
  
  
  
  
• start the ec2 instance
• Horizontal scaling:
  • This can be done manually by adding ec2 instances with the same AMI.
• Consider the below architecture
  
• As number of parallel users increase, load on web, orders and catalog increases which might lead to performance bottleneck.
• During the peaks or when in need we need aws to automatically increase number of ec2 instances running the corresponding application.
• It is better to create an AMI for
  • web
  • orders
  • catalog
• AWS has an option to perform auto scaling which is referred as Auto scaling group.

Auto Scaling Groups

• AWS EC2 Auto Scaling Groups give us two broader options
  • fixed scaling:
    • In Fixed scaling the fixed count is maintained by AWS
    • For eg: you have give fixed scaling count to 5, it creates 5 ec2 instances in chosen subnets (AZ’s). By any chance if the ec2 instance is down, aws will create one more to maintain count of 5
  • auto scaling:
    • In this case we specify
      • min ec2 instances
      • max ec2 instances
      • desired count
    • We also specify when the ec2 instance count should be increased (scaled out) and decrease (scaled in)
    • desired count will be dynamic depending on situation.
• To Create Autoscaling group
  • AMI: Create AMI with Applications pre installed
  • Decide which metrics to use when your ec2 has to be scaled out or in
  • Choose the AZ’s (subnets) where scaling has to happned
  • Choose the scaling methodologies

Experiments

• Create an ec2 instance with phpinfo and stress

#!/bin/bash
apt update
apt install apache2 stress -y
apt install php libapache2-mod-php php-mysql -y
echo '<?php phpinfo(); ?>' > /var/www/html/info.php
systemctl restart apache2

• Navigate to monitoring on AWS Console
  
• ssh into ec2 and execute

stress --cpu 10 --io 40 --vm-bytes 256M --timeout 10m -v

• observe the metrics on the AWS Console for next 10 minutes
  
• Generally in servers like apache the load on cpu increases as number of requests to websit increases, so cpu utilization should be a decent metric to start scaling