Design and Implementation of a Reusable Knowledge Model for Supporting the Network Management Functions

Sameera ABAR, Tetsuo KINOSHITA
2011 Interdisciplinary Information Sciences  
The focus of our work is the elicitation of communication network systems' knowledge resources in a generic and reusable manner for providing the automated support to network management tasks. Key features of the proposed knowledge model are: ontological representation of static domain-content and management-expertise encoded as the core knowledge of distributed multi-agent architecture. Our emphasis has primarily been on the modularization of resource knowledge to facilitate its reuse in a
more » ... ible manner. To demonstrate the effectiveness of proposed scheme, we have implemented an experimental network in our laboratory, and the devised knowledge model has been deployed through multi-agent based middleware layer in the prototype system. A couple of application scenarios have been designed for testing with the prototype system. Experimental results confirm a marked reduction in the workloads of the network operator with our system providing the automated support to network management functions. Validation of the reusability/modifiability aspects of our system illustrates the flexible manipulation of knowledge fragments within diverse application contexts. We envisage our knowledge modeling approach as the first step towards the comprehensive knowledge acquisition, representation, and dissemination in the communication network management domain. automation of management functions is the detailed interpretation of network-related information and knowledge resources. Traditionally, knowledge engineering was viewed as a process of "extracting" knowledge from a human expert and transferring it to the machine in computational form. Today, knowledge engineering is approached as a sophisticated modeling activity, and inference or reasoning algorithms are used to solve problems with the help of this knowledge. The applications are characterized by the tasks and domains involved. Knowledge modeling can, therefore, be divided into two conceptual sub-activities: (a) modeling the domain knowledge and (b) modeling the task knowledge (Abu-Hanna and Jansweijer, 1994) . The "Heuristic Classification" (Clancey, 1985) method relies on the experiential knowledge of systems and their behaviors. In contrast with other descriptions of expert systems, this method specifies the knowledge needed to solve a problem (through the useful combinations of problem solving tasks and associated sub-tasks for the purpose of sharing and reuse) independent of its representation in a particular computer language. Chandrasekaran, (1992) spotlighted the knowledge modeling approach in terms of the notion of a "task structure" which recursively links a task to alternative methods and to their sub-tasks. In the proposed work, the characterization of the network knowledge model has been performed in-line with the CommonKADS (Schreiber et al., 1994) -a methodology for expertise modeling embraces the application-intensive knowledge in three types as: domain, inference, and task hierarchy. Recently, Multi-Agent System (MAS) has emerged as a flexible way to manage the resources of distributed systems. In this paper, MAS-based approach has been adopted to deploy the functionality of the intelligent or autonomous behavior of proposed network knowledge model, to aid in the reduction of network management workload. Multi-agent systems are composed of multiple interacting agents where each agent is a coarse-grained computational system in its own right, as well as independently modifiable (Hamdi, 2006) . Agents, while being well-focused on their automated tasks, provide inherently distributed solutions. While the multi-agent research area is very active and it offers an appropriate tool to tackle the network-related problems, its concerns towards this domain are not yet well covered. The scant evidence of real-world deployment of agent-based systems is clearly being considered due to the same knowledge engineering bottleneck that has been the choke-point for the widespread application of expert and intelligent systems. At present, machines and software can store the information, rank it, display it, but cannot comprehend or process it. Trying to overcome this issue led to the basic notion that if we could be able to make this information inter-linked for sharing among the agents. Therefore, it is required to make a more schematic organization of information -an organization that is more suitable for machines to operate on for the provision of various automated services. The nextgeneration multi-agent based Knowledge Management (KM) systems will likely rely on conceptual models in the form of ontologies -the shared conceptualizations of a domain of interests, and these structure the knowledge resources in a highly expressive manner for the purpose of efficient reuse. KM facilitates the capture, deployment, access, and reuse of information typically using contemporary technology (O'Leary, 2001) . Ontologies: the keystone of new generation of multi-agent based systems pave the way to move from a documentoriented view of the KM to a content-oriented view, where knowledge items are structured, interlinked, combined and used, thereby facilitating agent interactions and communication with the sources. Ontologies have become ubiquitous in information systems (Noy and Musen, 2004) . Recently, we have seen an explosion of interest in ontologies as artifacts to represent human knowledge as critical components in knowledge management, the Semantic Web, e-commerce applications, bioinformatics and several other application areas (Brewster and O'Hara, 2004) . The proposed approach utilizes a concise and consistent ontology representing the static content of the communication network system domain, capable of fitting as more as possible the technical descriptions of objects (network-related terminologies). The proposed domain ontology design in the network management domain presents a hierarchical structure, which glues together classes representing network entities and association between them. Our key concept is to specialize the agent interactions with the network systems' knowledge resources, for autonomously and flexibly managing the network devices and resources, thereby reducing the workloads of a network administrator remarkably. Motivation Motivation for this research originated from the need to devise a MAS-mediated and ontology-driven knowledgebased strategy in support of the automatic provision of just-in-time and just-enough, context-dependent knowledge for actively managing the data communication network systems (Abar et al., 2005) . So far, a little work has been done for managing the operational knowledge of the communication network systems. Hence, the proposed idea can be regarded as an initial step towards the acquisition, representation, utilization, and sharing of widely distributed network knowledge resources. Another reason for devising our network knowledge representation scheme stems from the fact that not many knowledge modeling techniques have been developed for the diagnostic technical domains. The main draw-back of knowledge-based systems is a need for knowledge acquisition -a well-known bottleneck for many artificial intelligence applications. Building new knowledge-based systems today usually entails constructing fresh knowledge bases from scratch. It could instead be done by assembling reusable components (Neches et al., 1991) . Therefore, this work can serve as a test-bed to be reused for various practical diagnostic domains. The proposed approach embodies the 20 ABAR and KINOSHITA Design and Implementation of a Reusable Knowledge Model for Supporting the Network Management Functions 21
doi:10.4036/iis.2011.19 fatcat:44nzg5zbkbex3mtcn2wm2wx2bq