Auto Scaling Groups Overview

Instructor: So now let's talk about what is an auto scaling group. So when we deploy a website or an application, the load can change over time because we may have more users visiting our websites over time. And we've seen that in the cloud, in AWS, it's possible for us to create and also get rid of servers very quickly with the EC2 instance creation API call. So, if you wanna automate this, we can create an auto-scaling group. So the goal of the ASG is to scale out, that means add EC2 instances and you need to remember this, scale out to match an increased load or to scale in that means removing EC2 instances to match a decreased load. And so therefore the size of our ASG is going to vary over time. Overall, we can also define parameters to ensure we'll have a minimum and a maximum number of EC2 instances running at any time in our ASG. ASG also have the superpower that if you are pairing it with a load balancer, then any EC2 instances as part of the ASG will be linked to the load balancer. Another superpower is that if one instance is deemed unhealthy, it is terminated and a new EC2 instance is created to replace it. So, auto scaling groups are free and you're only going to pay for whatever resources are created underneath such as your EC2 instances. So let's see how an ESG works in AWS. So we set a minimum capacity, which is how many instances you want at minimum in your ASG. For example, two. Then you set a desired capacity, which is how many instances you want in your ASG, for example, four and then you set a maximum capacity which is how many instances at a maximum do I want in my ASG. And then that means that if you move the desire capacity to higher number, but that's still less than a maximum capacity, then you can scale out as needed. That means scale out means adding EC2 instances and so therefore your ASG can grow bigger and bigger and bigger. In this instance, the maximum capacity is seven. As I said, the ASG also works with a load balancer. So if we have four instances registered in our ASG, then the ELB is going to distribute traffic to all these instances right away and so your users can access a load balanced website. But the ELB also has the ability to check for the health of your EC2 instances using the health check and that health check can be passed on to the ASG. That means that the ASG can terminate EC2 instances if they are deemed unhealthy by the load balancer, which is very handy. Also, if you scale out, that means if you add EC2 instances, then of course the ELB is going to send traffic to them as well and spread the load. Therefore, it's a really great combination to use a load balancer and an auto scaling group. Now, in terms of attributes to create your ASG, you need to create a launch template. There used to be a thing called launch configurations but it's deprecated, but the idea is the same. A launch template contains information on how to launch EC2 instances within your ASG. So you have information about the AMI and the instance type, the EC2 user data, the EBS volumes, security groups, SSH key pair, IAM roles for your EC2 instances, network and subnet information as well as load balancer information and more if you want. And so all these parameters look a lot like the ones we specified when we created an EC2 instance. On top of it, your ASG has a min size, a max size and an initial capacity we need to define as well as scaling policies. Talking about scaling policies, let's see how CloudWatch alarm integrates with auto scaling. So you don't know what CloudWatch is yet of course, but let me just tell you a little bit about it right now. So it's possible to scale in and out, an ASG based on CloudWatch alarms. So for example, we have an ASG right here with three EC2 instances and the alarm is going to be triggered and therefore we're going to get a scale out activity. So what would trigger an alarm? Well, it's a metric you can return, for example, for the average CPU or any custom metric you want. So for example, if the average CPU as a whole for your ASG is too high, then you need to add EC2 instances and therefore the alarm is going to be triggered and is going to trigger a scaling activity in your auto scaling group and this is why it's called an auto scaling group because paired with alarms, there is an automatic aspect of scaling behind the scenes. So based on the alarm, we can create scale out policies. That means increasing the number of instances or we can create scale in policies to decrease the number of instances. And all these things together are what composes ASG. I hope you liked it and I will see you in the next lecture.