Buseung Cho

CBR-based network performance management with multi-agent approach

Nowadays, with the growth of heterogeneity of network equipment and software, and involved a variety of advanced network technology, the complexity of computer network have rapidly increased. Especially demands for high-performance network service in advanced data-intensive scientific research are dramatically increased. In order to cope with the network performance issue, this paper proposes the end-to-end (ETE) network performance management framework based on case-based reasoning with the case library and multi-agent integrated with perfSONAR as well as large-scale network flow monitoring. It provides a sophisticated framework for both network operator and user to systemically identify ETE network performance issues that are detected, diagnosed and recovered. The real cases are modeled and implemented into the casebase in the real experimental environment, a national research network of Korea. To verify the proposed framework, validation of the cases in the casebase is demonstrated.

End-to-End Network Performance Management Framework Based on Case-Based Reasoning

In the high-throughput scientific applications or researches, the end-to-end high performance data transferring is a critical issue in network and end-systems. If the performance of data transfer were not guaranteed, their researches couldnt achieve successful results and even more the researches couldnt start to do. In order to cope with the complex fault of the end-to-end transmission performance, we proposed the end-to-end network performance management framework based on case-based reasoning with the case library and multi-agent integrated with perfSONAR. We validated the case retrieving process with similarity measure and the real case that results from the performance degradation in the campus backbone segment.

A Fast Algorithm for Generating Virtual Dedicate Network Based on Software-Defined Wide Area Network

Various needs of new network services have rapidly proliferated over the past few years. To catch up with major requirements, that is, time-to-research and time-to-collaboration, we propose a fast virtual dedicate network generation algorithm based on network abstraction with pruning strategy, unification of multiple links, and an improved spanning tree algorithm.

Designing of Healthcare Information Network Using IPv6 Cloud Networks

Information networks have emerged a major research area in the interest of healthcare. There are several applications in the R&D field of healthcare. Healthcare Network must ensure the reliability and efficiency because it transfer data of health. We consider distributed cloud systems, which deploy IPv6 and agents that are geographically distributed over a large number of locations in a wide-area network. In this article, we also argue for a healthcare information management model that provides higher-level connectivity and logical network abstraction that are integral parts of wellness applications.

The Isolation Algorithm of Problem Location with Multi-agent Approach for End-to-End Network Performance Management

With the dramatic increase in demands for high-performance network service, practically in advanced scientific research, as well as commercial internet service, the performance of data transfer for the huge amount of scientific experimental data and medical data, genome data is the key element that enables advanced global collaborative research. These activities require a guaranteed end-to-end (ETE) network performance and sophisticated network management framework for the ETE network. This paper suggests an isolation algorithm of problem location with multi-agent approach isolation algorithm of problem location with multi-agent approach ETE network performance management framework with Case-Base Reasoning (CBR) approach and multi-agent approach. It will enable a preliminary and proactive performance management for ETE network.

Comparative analysis of darknet traffic characteristics between darknet sensors

Today, Internet is incessantly attacked by wide variety of network-based threats. One of the ways to monitor or identify such prevailing threats is to monitor incoming traffic to unused network addresses popularly known as darknet and often also referred with various other names like network telescope or black hole. As, all the traffic arriving at darknet is mainly the result from malicious probing or mis configuration in the network. It is expected that to have similar incoming traffic behaviour across different darknet sensors, however, various studies found it different. Various reason cited behind it is misconfiguration, certain kind of attack, difference in filtering parameter or system configuration itself. However, concrete reason beside this is still missing. In this regard, to get further understanding, in this study, we performed deeper comparative analysis between two darknet sensors (KISTI Darknet network) that are differently located but have similar filtering and system configuration. Comparative analysis considering total incoming packet, number of source host, targeting destination port and protocol revealed that there exists wide difference in incoming traffic characteristics between the darknet sensors. Moreover, for TCP and UDP comparison, UDP traffic showed more targeting behaviour to particular darknet block (difference in traffic characteristics between darknet sensors), in contrast to it, TCP traffic showed more scanning behaviour (similarity in traffic characteristics between darknet sensor).

Design of National Research Network for Strengthening of Supporting an Advanced Research Activities

Recently, a huge network backbone with 100Gbps or higher as well as 10Gbps has been developed and used in various R&D sectors. Based on this kind of a large-scale backbone network, international collaboration and sharing of a lot of research resources have become more common in R&D activities. It is very important to design and support an appropriate national R&D network.

International Research Networking of KRLight Supporting Data-Intensive Science

Recently global collaboration research related to High Energy Physics, Astronomy, Climate Change, etc. has been gradually increased based on international research networking with hybrid networking service. For more than 10 years, GLIF Open Lightpath Exchanges (GOLE) in Global LambdaGrid Integrated Facility (GLIF) with hybrid networking technology like lightpath is critical stack to transfer big e-Science data that support global collaborative research. KRLight has become a key GOLE in Asia-pacific area bridging research network between Asian research network and North American and European research network with dynamic circuit networking facility. Today emerging technologies such as Science DMZ, Software-Defined Networking, perfSONAR and GLORIAD Insight system could be leverage networking technology to support data-intensive science with more effective and flexible. In this paper we proposed new GOLE architecture to apply these network technologies into current GOLE architecture as a key international research networking infrastructure and validated it through adaptation into KRLight.

Designing a Specific Group Network Measurement on Light Path

Scientific co-work needs not only cut-edge network infrastructures for large-scale networks but also network technologies for high-performance networks. Specific group network is connected several sites over multi-domains It should support dedicated network resources and advanced measurement for collaborative scientists. In this paper, we introduce measurement technologies and apply network measurement framework to specific group over LP (Light Path) network for scientific communities.