For network service providers, assessing and monitoring network parameters according to a Service Level Agreement and optimal usage of resources is important. Packet loss is one of the main factors to be monitored, especially when IP networks carry multimedia applications. Measuring network parameters is more valuable when it is accurate and online. In this paper, the authors propose an accurate approximation for packet loss probability at an intermediate high speed node with finite buffer, where a large number of sources are expected to be aggregated. In this method, based on Large Deviation Theory, estimation of packet loss probability at the intermediate nodes is based on the input stochastic traffic process.