New method for predicting GridFTP throughput
Baoyi Wang, Shaomin Zhang, Qiaoli Xue · 2008
GridFTP is a secure, efficient data transfer protocol in grid, so GridFTP throughput is one of the important indicators for measuring the data transferring rate between grid nodes. In a grid, it is helpful to predict the GridFTP throughout between grid nodes for choosing efficient transformation and preventing congestion. A new method for predicting GridFTP throughput between grid nodes is presented in this paper. According to the chaos theory, firstly by means of phase space reconstruct the GridFTP throughput time series. After analyzing the characteristics of the grid and its transformation, adding weight one-rank local region method is adopted to predict the self-similar GridFTP throughput directly, which doesn’t consider the stochastic factor related to it. The simulated result indicates that this method has high accuracy, is real-time and can acquire better forecasting effect comparing with other methods. Predicting the GridFTP throughput based on chaos theory is seen as a significant attempt.