Hello and welcome to Vertical Scaling vs. Horizontal Scaling. In the next lesson, I'm going to perform a demonstration that shows you how to deploy a VM scale set that implements autoscale. However, before we do that, it makes sense to quickly explain to you the different ways that you can scale.

There are two different ways of scaling. You can scale up and down, or you can scale in and out. Scaling up and down, which is also referred to as vertical scaling, is the process of upgrading an existing virtual machine to a more powerful virtual machine, or downgrading to a smaller, less powerful virtual machine. In Azure, this means resizing the existing VM to a larger size with more RAM and more CPU, or to a smaller size with less RAM and less CPU. Because vertical scaling involves the re-size of virtual machines, this kind of scaling results in VM restarts.

Scaling out is a way to handle an increased load on an application or website. Scaling out is commonly referred to as horizontal scaling. When you scale out, you keep the same VM size, but you add more VM instances to the scale set. Instead of growing the size of the VMs in the scale set, you increase the number of VMs in the scale set. When you scale back in, you reduce the number of VMs in the scale set.

You might want to use vertical scaling to scale down an application's server over the weekend, when the application is in use. By scaling down over the weekend, you can save on compute costs.

A common use for scaling in and out, or horizontal scaling, is when you have a public-facing website or web application. When supporting this type of application, it usually makes sense to add instances to support increased workloads, while removing instances when that load drops. This allows you to keep up with demand without causing an outage.

Microsoft recommends horizontal scaling whenever possible.