I think it's worth noting (and please correct me if you think I'm wrong), that we may have multiple clusters. For example, our code instructions will obviously often cluster particularly when sequential execution, or looping occurs. And independently of this our execution stack data will also cluster for obvious reasons. (We really only need to know the things pushed onto the stack at the time of a function call IE it's parameters, return address and then anything defined locally to the function after that). This can lead to multiple clusters for different sets of information we use in a program. This doesn't really change how we can use the concept if you think about it. It's just worth keeping in mind.