A novel adaptive bandwidth allocation: wavelet-decomposed signal energy approach

Traffic measurements are key in the development of robust resource management and scheduling algorithms. Bandwidth allocation methods are either static or they adaptively change the bandwidth assigned to applications. Adaptive algorithms predict the near future behavior of traffic based on average queue size and utilization information to update the allocated bandwidth. This work introduces a novel feedback mechanism that uses wavelets and measures the energy of a byte counting process at the output of each dyadic subband filter. The output is then used to predict short and long term fluctuations in the incoming traffic pattern and to readjust the bandwidth allocation accordingly