Soft Computing Agents for Dynamic Routing

Georgi Kirov
2003 Information & Security An International Journal  
Introduction The strength of the distributed systems 1 as a new approach derives not only from its ability to allow people to communicate across big distances and at different times, but also from the ability of machines to help people communicate and manage information. The pace of change in the software industry is so great that the traditional distributed solutions may run out of steam in the not too distant future. The forces of competition, regulation, convergence and globalization are
more » ... ing change. Each one of these forces is changing the way we realize interconnected applications and the nature of networks and services. At present, the existing software technologies and network management infrastructures are getting old and information distributed systems cannot cope with the speed of response required in tomorrow's world. It is questionable whether traditional computing technologies can cope with the total information management demands of global network applications that companies will need to field in the early 21 st century in order to remain competitive. Increasingly, a great variety of different software applications and worldwide information services, such as WWW servers, databases, software packages, distance learning, video on demand, are being connected through the Internet, intranets, and other network systems. With the arrival of new technologies it is necessary to attempt to coordinate the action of these disparate entities within a cohesive framework. Standard distributed systems rely on message-based exchange with fixed connections, and as such, these systems are of limited efficiency when one attempts to use a large number of applications. Situations like these are prohibitively expensive in both time and resources, unless network applications cooperate effectively through an appropriate communication infrastructure. The field of distributed network systems is in a critical need of intuitive and innovative approaches and novel algorithms to Georgi Kirov 105 address the growing complexity in all of its different aspects: performance, stability, security, connectivity, efficiency, routing. Current lines of research give the promise of stopgap solutions that will suffice for the next five to ten years. Areas such as distributed artificial intelligence (agent technology) appear to offer good term solutions for distributed network applications and service management. Current engineering technologies could breathe a breath of fresh air into management information systems. But whilst these newer approaches will partner humans in dealing with the forthcoming explosion in scale and complexity, they only offer better, proactive, access to information stores and expert system solutions that we enjoy today. Soft computing could multiply the benefits of such systems many times. 2,3 Distributed information systems can be viewed as two level structures:  A network level that deals with network traffic, management, and control; and  A service level that deals with applications, inter-application communications, and service provided to the customers. Soft computing technologies have had an impact on these two levels to a varying degree. One of the most popular soft computing technologies, fuzzy logic, has been applied to the network level for network routing, traffic modeling, and congestion control. 4 The service level, the area of inter-application communications in particular, creates an opportunity to address problems within the Artificial Intelligence (AI) domain, e.g. within the intelligent distributed information systems and the intelligent multi-modal interfaces. Soft computing in conjunction with other AI techniques and software agents 5 can be used for knowledge representation and reasoning, information retrieval, search and optimization to make the resulting systems more robust, flexible and adaptive. In an attempt to resolve some of the above-mentioned problems in network communications the author proposes an approach that combines the Bee-gent agent technology with the fuzzy logic representation. The Bee-gent Technology Basic Concept Bee-gent is a communication framework based on the multi-agent model. 6 It has been developed by the Toshiba Corporation. It provides applications with autonomous 106 Soft Computing Agents for Dynamic Routing network behavior by "agentifying" them. Bee-gent supports agent-based interapplication communication, facilitating co-operation and problem solving. This environment is based on two types of agents -Agent Wrappers (AW) and Mediation Agents (MA). The main function of the agent wrappers is to agentify existing software applications, while the mediation agents are responsible for interapplication coordination by handling all communications. The MA can move from an application to another, interacting with the AW. The AW themselves manage the state of the applications they are wrapped around. The Bee-gent applications are suitable for many software fields: distributed databases, management systems, and system optimization. 7 Figure 1 illustrates the relationships between existing applications, the agent wrappers, and the mediation agents. 8
doi:10.11610/isij.1206 fatcat:cjuzfjsv4zdefei7uckoqq4gl4