Research Area: Software Defined Networks
Multipath networks have been extensively studied to improve throughput and alleviate network congestion. However, multipath networks, with the exception of equal-cost multi-path networks, are difficult to implement because they schedule network resources at cross layers and require complex signaling mechanisms to monitor network status (e.g., round trip time, bandwidths, and link state). Virtualization, in contrast, provides network isolation with protocols and resources and simplifies the process of coupling the interaction between signaling mechanisms, which makes multipath implementation practical. This paper proposes a multipath network virtualization scheme that implements software-defined networking (SDN) and network function virtualization (NFV). Multipath networks can be deployed in the virtualized environment. The proposed scheme attains a summary of network resources, such as network topology and link bandwidth, and can schedule these resources for selecting and spreading flow over multiple paths. In addition, virtualization provides computing and storage resources for flow splitting, tag adding, and packet reordering, among other applications. Specifically, the tags are designed for forwarding and packet reordering. Our experiment builds the platform based on open platform for NFV and SDN on the practice testbed. To verify its functionalities, we first evaluate the overhead of each component and then implement and compare five state-of-the-art multipath models on the proposed platform. Our scheme is compatible with these multipath models. Moreover, the experimental results show that the model with flow splitting has superior performance with respect to load balancing.
Keywords:
Multipath
virtualization
NFV
SDN
flow splitting
packet reordering
Author(s) Name: Qingtian Wang; Guochu Shou; Yaqiong Liu; Yihong Hu; Zhigang Guo; Wei Chang
Journal name: IEEE Access
Conferrence name:
Publisher name: IEEE
DOI: 10.1109/ACCESS.2018.2842058
Volume Information: ( Volume: 6) Page(s): 32460 - 32470
