Understanding performance bottlenecks of a multi-BSS software defined WiFi network testbed
Understanding performance bottlenecks of a multi-BSS software defined WiFi network testbed
复制标题
DOI：
10.1109/cci.2016.7778897
复制
发表时间：
2016-10
期刊：
影响因子：
--
通讯作者：
Tausif Zahid;Xiaojun Hei;W. Cheng
中科院分区：
文献类型：
--
作者：
Tausif Zahid;Xiaojun Hei;W. Cheng
WiFi网络由于其便宜但高带宽的基础设施而在过去十年中变得流行以提供网络接入。随着WiFi部署的日益密集以及网络设备和应用的快速增长，对更多带宽、受限延迟性能和更低功耗的需求正在飙升;然而，传统的WiFi网络已经无法解决诸如不平衡的业务负载、切换和物理信道变化等各种挑战。在具有多个接入点（AP）的WiFi网络中，传统的客户端与AP的关联通常基于分布式决策（诸如接收信号强度、最早关联应答等）来完成。由于AP之间缺乏协调，WiFi网络可能遭受不平衡的业务负载。新兴的软件定义WiFi网络（SDWN）引入了集中式控制器，它将控制平面和数据平面分开，也可以充当AP之间的自然协调器。在本文中，我们进行了测量研究，以了解SDWN测试平台的性能瓶颈。我们的研究结果表明，SDWN显示了巨大的潜力，灵活的网络管理和控制的性能下降的代价。我们基于Odin项目构建了这个SDWN测试平台。在这项研究中的性能指标检查集中在吞吐量和切换延迟时，移动的设备在AP之间漫游。在SDWN中创建的虚拟接入点有效地隐藏了AP之间的移动的设备的切换延迟;然而，与传统WiFi网络相比，吞吐量性能下降了约50%。我们的研究结果表明，目前的SDWN基础设施存在很大的设计和优化空间。
WiFi networks have becoming popular in the past decade to provide network access due to its inexpensive but high-bandwidth infrastructure. With the increasingly dense WiFi deployment and the rapid proliferation of network devices and applications, the demand for more bandwidth, constrained delay performance and less power consumption is soaring; however, the traditional WiFi networks have become incapable to address various challenges such as unbalanced traffic load, handover and physical channel variation. In a WiFi network with multiple access points (AP), the traditional client association with APs is usually accomplished based on the distributed decisions such as the received signal strength, the earliest association reply etc. Lacking of coordination between APs, a WiFi network may suffer from unbalance traffic load. The emerging software defined WiFi networking (SDWN) introduces a centralized controller, which separates the control plane and the data plane, may also serve as a natural coordinator between APs. In this paper, we conducted a measurement study to understand the performance bottlenecks of a SDWN testbed. Our results demonstrate that SDWN has shown great potentials of flexible network management and control at the cost of performance degradation. We constructed this SDWN testbed based on the Odin project. The performance metrics examined in this study focus on the throughput and handover delay when mobile devices roam between APs. The virtual access points created in the SDWN effectively hide the handover delay for the mobile devices between APs; however, the throughput performance degrades by around 50% compared with the traditional WiFi networks. Our results show that there exists large design and optimization space for the current SDWN infrastructure.