This computational model simulates the behavior of an "upstream" which is being fed requests at random intervals by a load balancer, and taking a random amount of time to process them. Each upstream has a maximum concurrency of 12 before queueing latency is incurred.
See this blog post to understand how this model can be used.