The previous two patterns show a very basic understanding of passing messages around a complex system, so that components (machines) can work independently from each other. While they are a good starting place, the system as a whole could improve if it were more autonomous. Given the previous example, we could very easily duplicate the worker instance if either one of the SQS queues grew large, but using the Amazon-provided CloudWatch service we can automate this process. Using CloudWatch similar to the Scale out pattern from Chapter 2, Basic Patterns, we might end up with a system that resembles the following diagram:

For this pattern, we will not start from scratch but directly from the previous priority queuing pattern. The major difference between the previous diagram and the diagram displayed in the priority queuing pattern is the addition of a CloudWatch alarm on the myinstance-tosolve-priority queue, and the addition of an auto scaling group for the worker instances...