Network performance optimization using importance sampling theory

In simulation runs which involve rare events, importance sampling (IS) is often used to speed up the simulation process to get the simulation result faster. This paper proposes a new application for the importance sampling theory: to predict the change in the network behavior when the network input changes. This paper will show, by modifying the way the importance sampling theory applied in rare event simulation, it is possible to calculate precisely the amount of adjustment required in the network input for the network performance to meet a predefined target.