A study on FARIMA-based dynamic bandwidth allocation for self-similar traffic

摘要

Recent measurements of LAN or ATM network traffic have shown that network traffic exhibits self-similarity. Analytical and empirical studies have shown that self-similar traffic can have a detrimental impact on network performance, such as packet loss probability. Furthermore, it has reported that buffering and traffic shaping techniques have little effect on reducing the above impact. In this paper, we study oil FARIMA (Fractional AutoRegressive Integrate Moving Average)-based dynamic bandwidth allocation for ATM switch. FARIMA, which can model asymptotically self-similar traffic, enable the prediction of arrival self-similar traffic. We compare the FARIMA-based scheme, with ARMA-based dynamic bandwidth allocation and constant bandwidth allocation in terms of CLP (Cell Loss Probability).