Measuring And Optimizing The Latency Of Wireless Local Area Network

In recent years,wireless data traffic has witnessed an exponential rise due to the explosion of smart devices.Among these wireless networks,WiFi has served a dominant fraction of today's wireless traffics.In 2020,49%of the Internet traffic will traverse WiFi as the last hop[1].However,many popular interactive applications(Web browsing,video chatting,instant messaging,etc.)are very sensitive to latency.One key observation by recent studies is that WiFi can be the weakest link of round trip network latency in the wild[2].Therefore it is very important to reduce the latency of WiFi.There are two critical metrics of WiFi latency:WiFi hop latency and WiFi connection set-up time.WiFi hop latency is the packet level latency for data transferring between the WiFi client and AP(access point).WiFi connection set-up time is the time span between the time a user clicks the SSID(service set identifier)name of the AP s/he wishes to connect and the time his/her device obtains the IP address.There are three challenges which contribute to the high WiFi latency:the complexity and dyna,mics of WiFi surroundings,Inefficiency and not expandable of WiFi protocol for dense WiFi environment,Diverse implementations of WiFi modules for heterogeneous WiFi devices.In order to address the above challenges,we carry out the following studies in the aspect of AP,protocol and client.The contributions of the dissertation can be summarized as follows:(1)The dissertation has proposed the design and implementation of WiLy to help understand,troubleshoot and optimize the WiFi hop latency for WiFi APs in general.To help understand the complex and dynamic WiFi environment,this paper has proposed an effective approach to break down the round-trip time into different parts:downlink latency,uplink latency and wired network latency.Based on the measurements and analyses,this paper has trained a general decision tree model to help the AP owner diagnosing the high WiFi hop latency events.The evaluation in the wild shows that WiLy can effectively reduce the WiFi hop latency by providing the guidance of how to re-configure the problematic APs.(2)The dissertation has proposed the design and implementation of WING to measure the interference of neighboring APs on WiFi downlink latency.It is critical to understand and minimize the WiFi interference in order to reduce the WiFi hop latency.For the first time in the literature,this paper can quantify the impact of interference on each actual packet without the assistance of extra hardware and tight time synchroniza-tions to maximize the deployment opportunities.By distinguishing different neighboring interfering APs,WING can also measure the time-variant interference graph between different APs.Such interference metric and interference graph enable the optimization on single AP and multiple APs for minimizing the interference and improving performance respectively.(3)The dissertation has proposed the design and implementation of QAir:an amendment of WiFi CSMA/CA mechanism to provide an implicit priority for delay sensitive application based on the traffic volume.The IEEE 802.11 standard uses CSMA/CA mechanism to access the channel,which leads to the linearly increasing of per-frame latency with the number of contenders.QAir has a practical solution to control the contention level of the channel and assigns implicit priority using the traffic volume of different flows.Through real experiments,QAir can significantly reduce WiFi latency of delay-sensitive flows without sacrificing the network throughput.(4)The dissertation has conducted the first large-scale measurements in the wild and has proposed the algorithm to reduce the WiFi connection set-up time cost.WiFi connection set-up time cost is a critical metric which can affect the user experience on WiFi.Thus it is important for research communities to pay more attentions.This paper uses 0.4 billion WiFi sessions collected from a mobile App that tops the Android App market to help better understanding the connection set-up procedure and locating bottlenecks.Based on the measurement analysis,this paper has developed a machine learning based AP selection strategy that can significantly improve WiFi connection set-up performance including reducing the connection set-up failures and the cost of connection set-up processes.