Certainly, we want efficient uses of expensive computing resources. However, when those expensive resources are available, to not be able to use them while at the same time our users, whom we serve, are waiting for an application to respond is selfish on our part, and actually quite preposterous!

When a user is waiting for the application to respond and there are available computing resources, then parallelism is being limited and serialization has raised its ugly head. Serialization is death to performance maximization!

While related, serialization and queuing are distinct, and you don't want to confuse the two. Queuing occurs when a process waits. But a process can wait because of serialization issues, not just because all available resources are busy. Also, there are times when all efforts to parallelize a process have failed, resulting in idle resources while the process is running. Let's take a closer look at each of these situations.