Autoscaling is one of those key fundamental pieces that everyone will need for their elastic and fault-tolerant architecture. Well, at least the architectures that need servers and what have you. It is the glue that makes the well-oiled machine work. 

Autoscaling is also one of those AWS things that we take for granted. People just assume they will have it as part of their architectures, and sort of handwave it away to deal with at another time. They assume that someone else will set up the policies correctly, and everything will work as intended. However, autoscaling has a number of complexities that can make a huge difference in the quality and consistency of its performance; and that is what we are here to talk about today.

There are basically four different ways you can manipulate and modify the number of instances within an autoscaling group. 

You can Manually adjust the variables - modifying the top and bottom bounds, or the desired amount. 

You can Schedule scaling to happen at certain times of the day - adding or subtracting instances as needed. 

There is also Dynamic scaling, which is automatic and adds or removes instances as required. 

Finally, there is Predictive scaling, this uses machine learning to understand your average loads and provisions your instances based on training data.

Each of these methods have their pros and cons, and it's important to understand when and where they have their place.

Let’s start this lesson by looking over why you would want to manually scale your auto scaling groups.