Cost Effective Network Flow Measurement for Software Defined Networks: A Distributed Controller Scenario
Software-defined networking (SDN) has emerged as an evolutionary paradigm in datacenter networks by separating data from control plane and centralizing network decision making. Traffic flow measurement in SDN is relatively lightweight in comparison to the traditional methods. It enables flow measure...
Saved in:
| Main Authors: | , , , , |
|---|---|
| 格式: | Article |
| 语言: | English |
| 出版: |
IEEE
2018
|
| 主题: | |
| 在线阅读: | http://umpir.ump.edu.my/id/eprint/23635/1/Cost%20Effective%20Network%20Flow%20Measurement.pdf http://umpir.ump.edu.my/id/eprint/23635/ https://doi.org/10.1109/ACCESS.2017.2789281 https://doi.org/10.1109/ACCESS.2017.2789281 |
| 标签: |
添加标签
没有标签, 成为第一个标记此记录!
|
| 总结: | Software-defined networking (SDN) has emerged as an evolutionary paradigm in datacenter networks by separating data from control plane and centralizing network decision making. Traffic flow measurement in SDN is relatively lightweight in comparison to the traditional methods. It enables flow measurement system to overcome the issues of traditional measurement systems, such as cost and accuracy by employing a centralized controller. Nevertheless, a full physically centralized controller introduces negative impacts on the network as well as the measurement system (i.e., introducing extra overhead or accuracy issues). However, few efforts have been devoted to measurement techniques in SDN distributed controller architecture, where every controller pulls its corresponding flow statistics, and these statistics are required to expose by only one single expression as if they are collected by one controller. Moreover, the imposed costs of flow measurement in distributed controller architecture are still an issue that remains unsolved. In this paper, we attempt to fill in this gap and present a novel and a practical solution for a cost-effective measurement system in SDN distributed controller deployment. We also propose a synchronization mechanism for aggregating traffic statistics in the multiple controller model. We evaluate our method through extensive emulations in a datacenter topology and present our findings to demonstrate the impact of multiple controllers on overhead and accuracy. |
|---|