Software-Defined ScienceDMZ Construction using SDN/ScienceDMZ/Edge Computing for High Performance Big Data Transformation

Authors

  • Ki-Hyeon Kim
  • Dongkyun Kim
  • Yong-Hawn Kim

Abstract

Background/Objectives: Recently, research using big data and AI has emerged as major issue in the ICT field.However, as size of big data grows exponentially, it brings up problemsthat is slow data transmission speed and is unable to accommodate various network structures when transferring data on existing legacy network. Accordingly, researchers demand new network technology that applies dynamic, flexible and high speed network technology. Methods/Statistical analysis: It constructs network structure without network performance degradation by using ScienceDMZ technology that divide the network for research and the general Internet network and using security policy of high performance switch.In addition, KREONET-S infrastructure using SD-WAN technology for softwarization of KREONETis combined with ScienceDMZ to build flexible network that can be applied to various networks.When constructing network that combines the two technologies, the Software-defined SicenceDMZ (SD-SDMZ) is configured using Edge-Computing technology, which discards the centralized computing approach and configures system close to the user.SD-SDMZ technology combines ScienceDMZ and SDN technology to provide network suitable for user's needs through network programming andto transmit data at high speed.In addition, SD-SDMZ can solve the bottleneck problem caused by the large number of users by using Edge-Computing technology. Also Edge-Computing is effective in terms of user accessibility. Findings: SD-SDMZ technology combines ScienceDMZ and SDN technology to provide network suitable for user's needs through network programming andto transmit data at high speed.In addition, SD-SDMZ can solve the bottleneck problem caused by the large number of users by using Edge-Computing technology. Also Edge-Computing is effective in terms of user accessibility.Edge-Computing technology configure by installing Data Transfer Node (DTN) in KREONET regional network center and installing Kubernetes which is container orchestration technology. Kubernetes uses container-based virtualization technology to provide container-based AI computing environment. A new Kubernetes Container Network Interface (CNI) has been developed for network linkage between Kubernetes and KREONET-S controllers. The newly developed CNI is used network for organizing containers on DTN server. Improvements/Applications: In this paper, we measure performance of SD-SDMZ infrastructure using ScienceDMZ, SDN, and Edge-Computing. The first performance measurement measures performance of the newly developed CNI and compares it with other CNIs. The second performance measurement measures data transmission performance when data is transmitted using the established SD-SDMZ network. The second experiment confirms superiority of the SD-SDMZ infrastructure by comparing virtual machine, host machine and container environment.

Downloads

Published

2019-11-22

Issue

Section

Articles