Neural network application to traffic shaping architecture for the ESW broadband satellite system

EuroSkyWay (ESW) is a continental broadband data-communication satellite system which has the mission to provide connectivity services both to telecom operators and servi ce providers serving a wide users population with the appropriate levei of service. ESW architecture foresees a satellite constellation and it is organised into ESW sub-networks (ESW-SN) each of them being served by a singie satellite. An advanced traffic management architecture allows ESW to fully adapt the bandwidth of provided connections to diversified traffic supported. At this aim, ESW dynamic bandwidth allocation capability, allows to change dynamically the bandwidth of each connection even though ESW switching mode is based on classica) circuit switches. ESW traffic management architecture is distributed over three main functionai component: the Traffic Resource Manager (TRM), hosted on board, which provides the network side of the architecture, a traffic management agent which, from the ESW user terminaI side inter-operates with the TRM and, finally, the Usage Parameters Control (UPC) which provides the traffic shaping functionaIity within the ESW user terminaI itseif. The UPC component covers the set of actions devoted to monitor and contro) traffic offered for a given ESW connection, at the end-system accesso Its main purpose is to protect network resources from malicious as well as unintentionai misbehaviour, which can affect the QoS of other aiready estabIished connections, by deteeting vioiations of negotiated parameters and taking appropriate aetions. In partieuIar this component monitors the eonformanee of the traffic actually in input against the traffie descriptors negotiated at eonnection set-up time. In case of a traffic contraet conformanee vioiation, a proper policing action on the eells sent in excess shall be applied. A eonformance monitoring is provided in term of the following traffie deseriptor measurement: Peak Cell Rate, Sustainable Celi Rate and Maximum Burst Size. In order to meet the expectations of network users, in terms of parameters sueh as celI Ioss and eall bIocking probabiIity, the use of network resources must be carefully eontrolled. On the other hand, resources should be handied so that the network operator revenue is maximised over time. In generaI, two types of traffic contro I mechanisms are adopted: preventive traffie eontrol and reactive traffie eontroI. Neurai Networks (NNs) have severai properties that are vaIuabie when impiementing traffic contro I. NNs can implement direet adaptive contro I without an anaIytieai representation of the traffie but using a good representation ofthe probiem. Moreover, the distributed and parallel architecture ofNNs, ean be exploited in hardware implementations, largeIy availabie on market, which provide short and predietabie response times. These properties have addressed a neural network solution for the traffic management functionality Iocated in the ESW user terminaI. In this paper, a study of the traffie shaping funetionality and a dynamic resouree assignment mechanism for the ESW broadband satellite network is presented. A neurai network architeeture based on the feed-forward scheme is proposed in the resource management mechanism for an adaptive ondemand traffie resourees assignment using the Usage Parameters Contro I function.