An Agent-based Data Analytics Support Tool for Network Management Intelligence

Kazuto Sasai, Yusuke Tanimura, Hideyuki Takahashi, Gen Kitagata, Tetsuo Kinoshita
2017 International Journal of Energy Information and Communications  
Data analytics technics are attracted to the field of network management. However, since a vast amount of data is produced from the systems, it is difficult to analyze effectively in ordinal network administrative office. In this paper, we propose an agentbased approach to support network and systems data analysis. Using the functions of agent-based systems to interact each other, collaborating with an agent-base network management system, it makes the data analysis tasks to be easy. Actual
more » ... analytics experiment on the network management task, we found that the agent-based analytics tool can effectively support the novice network and systems administrators. 52 Copyright ⓒ 2017 SERSC continuously provide new methodologies to understand the data collected from the network systems, it is difficult to deploy new methodologies to the autonomic systems. Therefore, a new kind of autonomic network management foundation on which human developers and maintainers easily build, reuse, and exploit knowledge resources to the network management systems. Recently, Kephart [18] asserted that a view of decision-making systems should include both of machines and humans, and they collaborate each other to solve the problems, and they called that symbiotic cognitive computing. In this paper, as a foundation for symbiotic cognitive computing, we propose a function of supporting network and systems data analytics collaborating with our data-centric network management system, called Active Information Resource-based Network Management System (AIR-NMS) [19] . AIR-NMS is a set of active information resources (AIR) for network and systems management, which consists of individual information resource including the autonomic features to realize self-management [20] . We introduce an agent-based design of network data analysis tools that are autonomously organized by according to the request of users. It is expected that highly reusable and deployable symbiotic cognitive systems can be established by connecting agent-based network data analytics tools and AIR-NMS. Figure 1 shows the conceptual diagram of the concept proposed in this paper. In Section 2, we briefly introduce about AIR-NMS and agent-based analysis function for network and systems management. Section 3 includes the practical design scheme and a prototypical implementation. Section 4 shows some experiments which were conducted to evaluate the prototype system. Section 5 shows the conclusion of this paper. Figure 1. Conceptual Diagram of the Proposal Agent-based Analytics for Network Management Agent-based network management is a concept that the agents perform the tasks of network management instead of the human administrators, and the complexity of the network management tasks is reduced by the collaboration between the agents [21, 22] . On the other hand, the software agent technologies are applied to data analysis to scale down the complexity of the information analysis process by applying self-organization as conceptually introduced in [23]. Even so, since the origin of raw data such as detectors and these analysis tools is separated and dealt out, it is difficult to deploy new analysis methodologies into on-going analysis systems. As reviewed in [23], a data mining process
doi:10.14257/ijeic.2017.8.1.05 fatcat:bhthsm6fxfd4lckzyharoejojm