In many cases with layered architectures, a problem at one layer shows up first at a higher layer. So, a report of an unreachable service at the application layer may be caused by a problem delivering packets at the network layer (perhaps because of bad routing table information).
Think of how often a failure in a complex device like an automobile is traceable to a small fundamental unit of the whole, such as the oil reservoir or a loose wire. A small change at this micro-level, like a change to cable on a network, can have an enormous effect on the “layers” above like the engine and steering, or a network server. So much so that in many cases, it's best to start at the bottom and work your way up.
Common causes of network failure
Studies have shown that network failures are most often caused by one of the following:
The point is not only that these things can go wrong (and if unsteady power causes the node to fail, which category does it belong in?), but that the greatest risk to a network is along the paths of the links that make up the network, which is why it is common today to deploy routers and other key pieces of network equipment in quads: four devices mesh-connected by six links. No single link or device failure in a quad ...