At this point, we have established that you need to assess each of the current resource capacities in your cluster, express this in terms of the number of free or spare ECS tasks your cluster currently can support, and then use the worst-case calculation (minimum value) to determine your overall current cluster capacity. Once you have this calculation, you need to decide whether or not you should scale out the cluster, or leave the current cluster capacity unchanged. Of course, you also need to decide when to scale in the cluster, however we will discuss that topic separately soon.
For now, we will focus on whether or not we should scale out the cluster (that is, add capacity), as this is the simpler scenario to evaluate. The rule here is that at a minimum, you should scale out your cluster whenever your current cluster capacity is less than one:
In other words, if you do not currently have sufficient capacity in your cluster to support one more worst-case scenario ECS task, you should add a new instance to the ECS cluster. This makes sense, in that you are attempting to ensure that your cluster always has sufficient capacity for new ECS tasks as they are started. Of course, you can increase this threshold higher if you want more free capacity, which might be applicable for more dynamic environments where containers are often spinning up and down.