Quokka: Latency-Aware Middlebox Scheduling with dynamic resource allocation

详细信息    查看全文

• 作者：Qing Li^a ; Yong Jiang^a ; ^{jiangy@sz.tsinghua.edu.cn} ; Pengfei Duan^a ; Mingwei Xu^b ; Xi Xiao^a
• 关键词：Software defined networks ; Network function virtualization ; Middlebox
• 刊名：Journal of Network and Computer Applications
• 出版年：2017
• 出版时间：15 January 2017
• 年：2017
• 卷：78
• 期：Complete
• 页码：253-266
• 全文大小：1367 K
• 卷排序：78

文摘

In the current Internet, middlebox management has become a significant challenge for network operators. Some schemes based on Software-Defined Networking (SDN) have been proposed in academia to simplify middlebox scheduling. However, these schemes are inefficient with the dynamical traffic requirements, as they mainly focus on stationary hardware middleboxes. Furthermore, the latency cannot be guaranteed in these schemes. In this paper, motivated by Network Function Virtualization (NFV), we propose the scheme of Quokka to solve these problems. In Quokka, we build a management framework for both computing resource and flow latency. We also design a latency model to describe the latency behaviour of flows. Based on the framework and model, we present a latency-aware scheme Quokka with portable software-based middleboxes that can be dynamically scheduled (placed) according to the changing traffic and resources. Quokka controls the latency by dynamically positioning the software middleboxes and scheduling the changing traffic. Therefore, different IT services with different latency requirements can be satisfied in Quokka. Comprehensive experiments show: (1) compared with traditional configuration methods, Quokka reduces the latency by about 20% on average; (2) Quokka requires 30–50% less middleboxes than traditional schemes to achieve the same performance.