Building a flexible and scalable virtual hardware data plane

Authors:
Junjie Liu;Yingke Xie;Gaogang Xie;Layong Luo;Fuxing Zhang;Xiaolong Wu;Qingsong Ning;Hongtao Guan
Affiliations:
Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China,Graduate University of Chinese Academy of Sciences, Beijing, China;Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China;Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China;Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China,Graduate University of Chinese Academy of Sciences, Beijing, China;Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China,Graduate University of Chinese Academy of Sciences, Beijing, China;Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China,Graduate University of Chinese Academy of Sciences, Beijing, China;Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China,Graduate University of Chinese Academy of Sciences, Beijing, China;Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China
Venue:
IFIP'12 Proceedings of the 11th international IFIP TC 6 conference on Networking - Volume Part I
Year:
2012

Citing 17
Cited 0

Overcoming the Internet Impasse through Virtualization

Computer
High Performance Switches and Routers

High Performance Switches and Routers
Network Virtualization: Breaking the Performance Barrier

Queue - Virtualization
OpenFlow: enabling innovation in campus networks

ACM SIGCOMM Computer Communication Review
Network Virtualization: A Viable Path Towards the Future Internet

Wireless Personal Communications: An International Journal
Towards high-quality I/O virtualization

SYSTOR '09 Proceedings of SYSTOR 2009: The Israeli Experimental Systems Conference
Towards high performance virtual routers on commodity hardware

CoNEXT '08 Proceedings of the 2008 ACM CoNEXT Conference
Trellis: a platform for building flexible, fast virtual networks on commodity hardware

CoNEXT '08 Proceedings of the 2008 ACM CoNEXT Conference
Network virtualization: state of the art and research challenges

IEEE Communications Magazine
Scalable network virtualization using FPGAs

Proceedings of the 18th annual ACM/SIGDA international symposium on Field programmable gate arrays
A survey of network virtualization

Computer Networks: The International Journal of Computer and Telecommunications Networking
SwitchBlade: a platform for rapid deployment of network protocols on programmable hardware

Proceedings of the ACM SIGCOMM 2010 conference
PacketShader: a GPU-accelerated software router

Proceedings of the ACM SIGCOMM 2010 conference
Network I/O fairness in virtual machines

Proceedings of the second ACM SIGCOMM workshop on Virtualized infrastructure systems and architectures
Customizing virtual networks with partial FPGA reconfiguration

ACM SIGCOMM Computer Communication Review
Network virtualization substrate with parallelized data plane

Computer Communications
VIOLIN: virtual internetworking on overlay infrastructure

ISPA'04 Proceedings of the Second international conference on Parallel and Distributed Processing and Applications

Quantified Score

Hi-index	0.00

Visualization

Abstract

Network virtualization which enables the coexistence of multiple networks in shared infrastructure adds extra requirements on data plane of router. Software based virtual data plane is inferior in performance, whereas, hardware based virtual data plane is hard to achieve flexibility and scalability. In this paper, using FPGA (Field Program Gate Array) and TCAM (Ternary Content Addressable Memory), we design and implement a virtual hardware data plane achieving high performance, flexibility and scalability simultaneously. The data plane uses a 5-stage pipeline design. The procedure of packet processing is unified with TCAM based rule matching and action based packet processing. The hardware data plane can be easily configured to support multiple VDP (Virtual Data Plane) instances. And in each VDP instance, the pattern of packet processing can be flexibly configured. Also, it can achieve seamless migration of VDP instance between software and hardware. The hardware data plane also provides a 4-channel high-performance DMA engine which largely reduces packet acquisition overhead on software. So that software can be more involved in customized packet processing.