Coping with broadband traffic uncertainties: statistical uncertainty, fuzziness, neural networks

Broadband networks will carry traffic with highly unpredictable traffic characteristics. Three approaches for dealing with traffic uncertainty are discussed and compared. The first is based on analyzing the effects of statistical uncertainty in traffic characteristics on queuing system performance. The second uses the theory of fuzzy sets in handling an uncertain service time of a queuing system, and in forecasting new services. The last uses a neural net approach to learn about the certain environment.<>