In general, data exchanged on an industrial network can be classified into two groups: realtime and non-realtime data. Non-real-time data do not have stringent time limits on their communication delays experienced during the data exchange. In contrast, real-time data have very strict time limits and the data’s value is diminished greatly as the communication delay grows larger. Therefore, when building an industrial network, the designer must configure the network to satisfy these requirements. MPLS provides extensive support for both integrated services/RSVP, and diff-serv QoS classes. Service providers can use MPLS to define classes of service (gold, premium, best-effort, etc.) and to define per-hop behavior (PHB) for each class to support the service. We use OPNET Modeler to design MPLS capabilities incorporating IntServ/DiffServ mechanisms into on an industrial network to ensure QoS under network failure conditions. We demonstrate how modern MPLS with QoS management techniques can control, but also complicate, prediction, and will finally illustrate how semi-empirical statistical techniques offer some resolution.