Our administrator course *previously* used small
instances, and we updated to medium instances because
the number of services and memory pressure was too much
for small instances.
We use VPC because the t2.medium instance size was the only 2 vCPU 4GB instance type available that supports an AMI of virtualization type HVM (which is how the image imported to EC2). Instances of the AMI can be deployed without the use of elastic IPs however, simply by configuring the VPC subnet to which you deploy to auto-assign public IP addresses to instances deployed within them.