A new regenerative estimator for effective bandwidth prediction

Motivation: Lately computer networks become more and more widespread. Sometimes a problem of guaranteeing of reliable and effective data transmission is critically important. In this connection it is essential to focus attention on such events, that can entail faulty operation or become a reason of loss of a message. One of such events is a great load of network components (for example, big queues of packets in units of a network). One can solve this problem using superfluous resources. But it is clear, that this method is economically unprofitable. At the same time insufficient resources can't provide demanded parameters of work of a network. Problem: What minimal volume of resources is able to provide given quality of service? There is also inverse problem: what quality of service can be guaranteed using given volume of resources? Theoretical base: Typically the load of network units is moderate. So it's drastic increase is a large deviation from typical values. So our research is based on the Large Deviation Theory (LDT).