Development in Handling the Problem of Fault Recovery in MPLS Networks Using Neural Network

Rapidly increasing volume of traffic carried by the Internet together with imposing requirements for reliability, quality of service (QoS) and manageability, force the network technology designers to come up with new approaches and solutions. As the Internet moves towards an IP over WDM model, existing means of network engineering to provide assured bandwidth, quality of service and fault tolerance should be substituted. Multiprotocol label Switching (MPLS) has emerged as a technology that can provide many of the functionalities now associated with ATM and/or SONET / SDH without incurring much of the overhead. MPLS appears to be a suitable place to provide fault tolerance. It is the lowest layer with the knowledge of the entire network topology as well as a point with the necessary traffic engineering capabilities. MPLS recovery mechanisms are applying in data communication network because they can guarantee fast restoration and high QoS assurance. Their main advantages is that their backup paths are established in advance, before a failure event takes place. Most researches on the establishment of primary and backup paths have focused on minimizing the added capacity required by the backup paths in the network. In this paper several back up paths are considered for a primary path. When fault occurs, regarding to remain bandwith of backup paths, the traffic of primary path is splitted over the backup paths using neural network(NN). Ultimately the performance comparision between CBR(case-based reasoning) and NN is presented.