Guest editorial: intelligent techniques in high speed networks

WE ARE pleased to bring to you this timely issue on the applications of intelligent techniques in high speed networks. Three years ago, the February 1997 issue was dedicated to the same topic. Then, the area of applying intelligent tools to high speed networks was relatively in its infancy. The objective of that issue was to convey to the reader the capabilities of intelligent techniques in solving many complicated problems that were challenging using conventional algorithmic methods. Tools such as neural networks, fuzzy logic, expert systems, and genetic algorithms, among others, were shown in many cases to outperform algorithmic methods. In some other cases, they were even the only “noncomplicated” feasible solutions. The dominant theme, then, was to tackle many of the ATM problematic traffic management issues such as admission control, policing, traffic shaping, and congestion control, among others. Many of the proposed solutions to these problems were simple and very successful in demonstrating their effectiveness either as “stand-alone” or in augmentation with conventional approaches. For example, multimedia traffic prediction for the purpose of traffic shaping or reserving resources, was shown to be much simpler and indeed more accurate using neural networks than traditional linear predictors. Similarly, extracting features and rules from incomplete data sets to compute bandwidth, or anticipate congestion episodes in switches, was simple and effective using neurofuzzy approaches. The uncertainties in estimating statistical parameters were also overcome through the use of several neural and fuzzy solutions. The advantages of intelligent techniques are numerous, most notably are learning from experience, scalability, adaptability, and the ability to extract rules without the need for detailed or precise mathematical modeling. In general, the advantages of pursuing this line of research have been proven worthy. Today, however, things have dramatically changed on the intelligent techniques front, as well as the networking front. First, intelligent tools do not need any further validation. They are not in their infancy any more. They are now relatively mature and are enjoying increasingly wider acceptance as many scientists and engineers are continuing to use them to tackle many of the emerging high speed networking problems. This is evident by the increasing number of publications, special issues in journals, technical sessions, and even some new products that all validate their effectiveness and usefulness. On the networking front, recent technological developments could not have been more exciting. Today, the dominant theme is integration. The integration of data, voice, and video traffic, the coexistence of ATM and IP, the seamless interoperability of wireless and wired services, and the interconnectivity of traditional circuit switching and packet