Zalal Uddin Mohammad Abusina is with the National Institute of Information and Communications Technology (NICT), Japan. He works for Japan Gigabit Network-II (JGN-II) Project at NICT’s Tohoku University Office housed in its Research Institute of Electrical Communications (RIEC). Salahuddin Muhammad Salim Zabir joined the Department of Computer Science and Engineering of Bangladesh University of Engineering and Technology in 1995. At present, he is with RIEC, Tohoku University. He is a member of the IEEE, BCS and BAAS. Ahmed Ashir received his PhD in 1999 from Tohoku University, Japan. He was with Japan Gigabit Network (JGN) Project of the Telecommu- nication Advancement Organization (TAO), Tohoku University Office. Debasish Chakraborty received his PhD in 1999 from Tohoku University, Japan. He is currently with Research Institute of Electrical Commu- nications, Tohoku University, Sendai, Japan. Takuo Suganuma is currently with Research Institute of Electrical Communications (RIEC), Tohoku University, Sendai, Japan. He is a member of the IEEE. Norio Shiratori is a professor at the Research Institute of Electrical Communication (RIEC), Tohoku University. He has been engaged in research on distributed processing systems and flexible intelligent networks. He is a Fellow of the IEEE, IEICE, IPSJ. *Correspondence to: Zalal Uddin Mohammad Abusina, Research Institute of Electrical Communication, Tohoku University 2-1-1, Katahira Aoba-ku, Sendai 980-8577, Japan. † E-mail: abusina@shiratori.riec.tohoku.ac.jp Copyright © 2005 John Wiley & Sons, Ltd. An engineering approach to dynamic prediction of network performance from application logs By Zalal Uddin Mohammad Abusina* ,† , Salahuddin Muhammad Salim Zabir, Ahmed Ashir, Debasish Chakraborty,Takuo Suganuma and Norio Shiratori Network measurement traces contain information regarding network behavior over the period of observation. Research carried out from different contexts shows predictions of network behavior can be made depending on network past history. Existing works on network performance prediction use a complicated stochastic modeling approach that extrapolates past data to yield a rough estimate of long-term future network performance. However, prediction of network performance in the immediate future is still an unresolved problem. In this paper, we address network performance prediction as an engineering problem. The main contribution of this paper is to predict network performance dynamically for the immediate future. Our proposal also considers the practical implication of prediction. Therefore, instead of following the conventional approach to predict one single value, we predict a range within which network performance may lie. This range is bounded by our two newly proposed indices, namely, Optimistic Network Performance Index (ONPI) and Robust Network Performance Index (RNPI). Experiments carried out using one-year-long traffic traces between several pairs of real-life networks validate the usefulness of our model. Copyright © 2005 John Wiley & Sons, Ltd. INTERNATIONAL JOURNAL OF NETWORK MANAGEMENT Int. J. Network Mgmt 2005; 15: 151–162 Published online 28 February 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nem.554