Look at f(x)=x3−3x. Near x=0 it is falling: f(−0.1)=0.299 and f(0.1)=−0.299. Near x=−1 it has a small peak, f(−1)=2, higher than every nearby value. Those are local facts. Now zoom out: f(10)=970 and f(−10)=−970. The peak of 2 is easily beaten further right, and the graph rises without bound on the right and falls without bound on the left. Those are global facts.
Zooming in makes almost any smooth curve look straight. Near x=0, sinx≈x: sin0.1≈0.0998. That is a local truth. But sinx never exceeds 1 while x keeps growing, so by x=3 the two are far apart, since sin3≈0.141. A local approximation describes one neighbourhood well and everything else badly, so it should never be read as a global formula.
Wording usually gives the scale away. Near x=2, at a point, on a small interval and relative maximum are local. For all x, absolute maximum, as x→∞ and the number of times the graph crosses the axis are global. A local maximum can also be the global one, as for −x2 at 0, but showing that takes a look at the whole domain.
Think of it like this. Local is the street-level view: you see whether your road climbs or dips and what is on this block. Global is the satellite view: the whole city, its highest hill, where the roads finally lead. Each view answers questions the other cannot, which is why a map app lets you zoom.
Core idea. Local behavior is what a function does near one point; global behavior is its overall shape across the whole domain.