“…Let us consider the feasible region of Example 1 ( Figure 4). Given an input rate combination p = (r 1 , r 2 ), maximizing δ 1 + δ 2 without violating constraints (2), (4) and (5), would result in a component rate assignment that is equivalent to the combination of application input rates p ′ = (r 1 + δ 1 , r 2 + δ 2 ), as shown in Figure 6. On the other hand, if a burst results in the application input rate combination given in Figure 7, we would prefer a plan that provisions for an application input rate combination p 2 = (r 1 + δ 1 , r 2 + δ ′ 2 ) rather than one that provisions for p ′ = (r 1 + δ 1 , r 2 + δ 2 ), since increasing δ 2 to δ ′ 2 will have no impact on δ 1 .…”