Throughout the previous chapters, services and infrastructure were described in the traditional fashion of front-end instances that users interact with, middle-tier instances that other systems (including the front-end) interact with, and possibly a back-end instance such as a database. A simple example can be visualized as follows:

We have described quite a few patterns that help to scale this type of infrastructure at each point, and all of them are valid here; but we can see a bottleneck in the API server. Let's presume that the system as-a-whole is working properly and all is well. Today, however, we have got an unexpected 500% increase in traffic to the front-end server.

We might scale the front-end instances but that would only move the bottleneck to the API instances. Again we could scale the API server, but perhaps the traffic is not utilizing S3 or Redis at all, and most of the traffic is going through to the RDS instance. Scaling out the API...