Yoo-Kyoung Lee

Capacity Planning with the Statistical Prediction of on-off Flows

A novel approach to describe the relation, in terms of probability, between the gain obtained by multiplexing the traffics in the packet network and the link capacity for guaranteeing the required quality is proposed. A traffic-flow is defined as the on-off stream described with the probability distribution of the on-off lengths. With the statistical on-off flow description, the loads on a link after aggregating traffic flows can be predicted in terms of probability, and the link capacity design based on the probability of supporting the required quality becomes possible. To have statistical model of a flow with small number of data, an interpolation method is suggested. To find a congestion point with given link capacity and traffic flows, a calculation method to estimate the expected link reservation with traffic flows and the congestion probability is proposed

Multiple spanning tree protocols in 10-Gb Ethernet edge system

The IEEE 802.1D STP is a kind of layer 2 link management protocol which provides path redundancy while preventing undesirable loops in the network. And it is based on all spanning tree protocols using an algorithm that calculates the best loop-free path through the network. The IEEE 802.1W RSTP provides rapid convergence comparing to 802.1D STP and the IEEE 802.1S MSTP allows frames assigned to different VLANs to follow separate paths, each based on an independent Multiple Spanning Tree Instance, within Multiple Spanning Tree Regions composed of LANs and or MST Bridges offering RSTP feature. These regions and the other Bridges and LANs are connected into one common spanning tree about three types of STPs.

On recovery algorithm for fault-tolerance in multicast trees

Since the multicast communication is the best technology to provide one to many communication, more and more service providers are using this technology to deliver the same service to multiple customers. These applications require seamless and real time services. With the deployment of the high-speed networks, real time services can be supported by reserving network resources in advance. In the case of seamless services, there should be no links or nodes failure in given networks. In real life networks, however, such failures are frequently happened more than we are expecting. In this paper, we propose a fault tolerant algorithm based on spanning trees that can be restorable using locally distributed mechanism in the case of multiple link-failures, if a tree existed. We also show that our algorithm can restore a multicast tree with constant recovery cost. We strongly believe our method can be generalized to apply to any type of tree-construction algorithm that requires the Quality of Service (QoS) in terms of reliability.

Ethernet-centric networking : Convergence, scalability, flexibility, and manageability

The Ethernet technoglogy rapidly gains importance as becoming a dominant solution for a converged transport network. Recently added features such as QoS handling and OAM functions in the framework of ITU-T NGN provides a means for performance improvement to meet carrier-class network requirements. In addition, integration with newly developed T-MPLS by ITU-T offers a potential to replace current SDH/SONET-based network infrasctructure.