期刊名称：Karbala International Journal of Modern Science
印刷版ISSN：2405-609X
电子版ISSN：2405-609X
出版年度：2018
卷号：4
期号：3
页码：313-331
DOI：10.1016/j.kijoms.2018.06.003
语种：English
出版社：Elsevier
摘要：AbstractThe emerged SDN innovation moves the intelligent routing functions out of forwarding devices into the centralized controller to provide the centralization and flexibility. It detaches the two planes (control and data) via its new architecture and programming solutions. Data centers benefit from SDN capabilities in the modern scalable network environments and tends to be managed by a central entity basing on its global view and intelligent dynamic decisions. Researches showed that elephant flows carried the largest amount of data in such data centers. These Large flows need to be detected and managed to enhance the resources utilization of the recent multi-path topologies.In this work, we developed the requirements for an efficient TE mechanism basing on an intelligent elephant flows management, and proposed the Sorted-GFF flow scheduler. Mininet simulation is used to calibrate the performance of the Sorted-GFF as compared to the GFF and the ECMP routing via examining them on a scalable low cost fat tree data center for various generated traffic patterns.
关键词：SDN;Load balancing;Elephant flow;Flow scheduling;Data center
