| Abstract |
This paper deals with an adaptive routing algorithm of self-similar traffic based upon the prediction of fractal time series and the GA (Genetic Algorithm). Mathematical results show that the superpos...ition of many ON/OFF sources exhibit the self-similarity. At first, the self-similar traffic is represented by using a convolution of the input signal and the impulse response function which is developed by using a set of scaling functions. The prediction method is derived by using the fact that the impulse response possess the self-similarity due to the fractal property even if the time scale is expanded. By using the fluid flow model where the input traffic is assumed to be fractal, the delay on the output link is predicted by our method. In the adaptive routing algorithm treated in our paper, the predicted delay on the output link is broadcasted as well as observed delay to find out shortest path in the next time point. By using the predicted delay and the observed delay to formalize routing tables, packets can be sent through more relevant path. Even more, by using these two kinds of delay, the time interval T for updates can be taken larger than in conventional algorithm. Conventional routing algorithms only generate the shortest path from the source to the destination, even if some alternative path is available. In the adaptive algorithm, we utilize the GA to find out some alternative path if the communication delay become large. The algorithm realizes a kind of load balancing by using the alternative paths. The performance of the adaptive algorithm is compared to conventional ones. As a result, the mean delay on the output link become about 20% smaller, and the variance of the delay become remarkably small.show more
|
Export Mendeley
108_p029