these are multi-layered systems of human, logistic, and operational complexity, that have multiple failure points that can cascade into notable failures for the end user.
this is because of emergent behaviors that arise only when multiple parts are moving at once.
in supercomputing maintenance, there are many components across datacenter, hostos, firmware, nvlinks, ib switches that need to be updated in tandem - one failure point in one component can compound when interacting with another point.
in healthcare, a cascade effect could come from icu bed management and nurse staffing, leading to unreliable service to patients.
in hardware or real-life factories, there are natural restraints to each component that deter failures. however these compounding failures are even harder to see in software - where each layer can move unbridled. in materials science, heatmaps demonstrate where stress concentrates in a structure, building up tension that can result in total collapse.
”One big misconception is that systems are broken and we need to fix them. In fact, systems are simply operating in the way they were designed to. Therefore, if we want them to stop producing a particular negative outcome, we need to focus our attention on redesigning some of the dynamics in the system (the feedback loops which form it) which are driving those outcomes. Systems maps help you see these dynamics and the consequences of a ‘design decision’ in a system.”
leverage points are where small efforts result in outside impact.