Top Banner
WiFox: Scaling WiFi Performance for Large Audience Environments Arpit Gupta, Jeongki Min and Injong Rhee NC State University
34

WiFox : Scaling WiFi Performance for Large Audience Environments

Feb 23, 2016

Download

Documents

olaf

WiFox : Scaling WiFi Performance for Large Audience Environments. Arpit Gupta , Jeongki Min and Injong Rhee NC State University. Interesting ??. WiFox : AP-only S/W solution. Faster WiFi !!. Manageable solution !!. Large Audience Environments (LAEs). Large Audience Environments. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript

WiFox: Scaling WiFi Performance for Large Audience Environments

WiFox: Scaling WiFi Performance for Large Audience EnvironmentsArpit Gupta, Jeongki Min and Injong RheeNC State University

Hi All,I am going to talk about my work WiFox which is about Scaling WiFi Performance for Large Audience Environments.

I collaborated with Jeongki and Dr. Injong for this work at NC State UniversitySource: WBA(Wireless Broadband Alliance)Interesting ??

Faster WiFi !!Manageable solution !!WiFox:AP-only S/W solutionWell before we proceed further we shall ask a question as whether this problem is interesting ??

Our Solution Wifox does exactly what both these communities desire and provide significant performance improvements as an AP-only S/W solution.

As a user of WiFi at public places like this conference or airports etc most of us had experienced poor performance and we desire to get better performance at such locations On the other hand vendors and SP desire to solve this problem in a more practical manner such that it does not requires changes to infrastructure or underlying protocols itself.

Source: WBA(Wireless Broadband Alliance)Large Audience Environments (LAEs)We start with a brief introduction to what we perceive as large audience environmentsSource: WBA(Wireless Broadband Alliance)Large Audience EnvironmentsAny location with a large WiFi user population WBA projected a growth rate of 350% per year for such WiFi deploymentsVarious successful deployment models such as Boingo-Google, Mobily-Aruba etc. already exist

Source: WBA(Wireless Broadband Alliance)Source: Wireless Broadband Alliance (WBA)So what is LAE and it is any location with large associated users population which involves places like malls, cafeteria etc. With the growing popularity of WiFi Offloading, Public Hotspots have gained significant importance in recent years. Wireless Brdbnd Allnce actually projected a growth rate of 350% per year for such WiFi deployments.

Source: WBA(Wireless Broadband Alliance)What about User Experience ?

As number of active clients increases, user experience diminishes significantlySo we understand the growing importance of WiFi for large audience environments but important question to ask is what about performance or user experience ?To an extent we know the answer to this question and that is as the number of active users increases performance degrades significantlySource: WBA(Wireless Broadband Alliance)Problem AnatomyWe now proceed to understand what causes this poor performance degradationSource: WBA(Wireless Broadband Alliance)Well Known Factors Contention and collisionIncreases with growing competitionRate DiversitySlower STA slows down all other STAsVarious fairness realizations like WFQ, TBR etc.Random Losses and TCP performanceTCP treats packet losses as congestion signalsUsage of TCP ECN and proxy servers isolate wired and wireless networksTraffic AsymmetryThere are various well known degradation factors and among them most intuitive one is Contention overhead . As number of active users increases the overhead for channel access increases and contributes to performance degradation.Then there is rate diversity which implies various STAs will have diverse link qualities and the ones with poorer link quality degrade the performance of entire network.. The most imp one is Traffic asymmetry.Source: WBA(Wireless Broadband Alliance)Traffic Asymmetry

Downlink Traffic dominates for 90% of data tracesBefore we define this asymmetry lets have look at few stats from network traces collected from SIGCOMM-08. The one presented here is the distribution of downlink to total traffic ratio and we observe that around 90 % of these traces had ratio greater than half.Source: WBA(Wireless Broadband Alliance)Traffic Asymmetry

Majority of data traffic is Web basedDownlink Traffic dominates compared to uplinkMajority of data packets are for HTTP based web activitiesWe also the analyse the application usage distribution and observe that most of the data traffic is HTTP based web activities. So for these L A E we make 2 imp observations observe that downlink traffic dominates compared to uplink and majority of these data packets are for HTTP based web activitiesSource: WBA(Wireless Broadband Alliance)Traffic AsymmetryIn scope of our problem it is:Downlink/Uplink AsymmetryChannel Access AsymmetryImplications:Packets spend more time at APs TxQFrequent packet dropsNow we will take a look at two factors that contribute to tis Traffic AsymmetrySource: WBA(Wireless Broadband Alliance)

Wireless ChannelUplink TrafficDownlink TrafficChannel access for uplink traffic is moreTo get a better idea about this problem we take a look at this abstraction of channel access for large audience environments. So we can note that Channel access is equally shared among associate clients and there are more packets for downlink than uplink. This implies that for existing access networks for large audience environments channel access for uplink traffic is more which contributes to performance degradationSource: WBA(Wireless Broadband Alliance)Performance Bottleneck

TxQ saturates as associated clients increaseAssociates UsersTo better understand the implication of higher channel access for uplink traffic we observe TxQ for WiFi AP for two different no. of associate clients.We observe saturation of transmission queue as number of active users increases. Source: WBA(Wireless Broadband Alliance)Goodput Performance

Traffic Asymmetry causes TxQ saturation resulting in poor goodput performanceWe now take a look at the resultant network goodput performance and observe significant degradation as number of clients increases. Based on these observations we can conclude that Traffic assymetry implies higher channel acces for uplink data traffic which results in performance bottleneck at the WiFi AP and consequently degraded goodput performance.Source: WBA(Wireless Broadband Alliance)Possible SolutionsNow we have an idea about the nature of problem and will now consider various possible solutionsSource: WBA(Wireless Broadband Alliance)

Wireless ChannelUplink TrafficDownlink TrafficEqual Channel Access for Uplink/DownlinkOne possible solution can be to provide equal channel access for both uplink and downlinkThough it solves the downlink/uplink asymmetry problem but does not takes into consideration that for most of the times downlink traffic dominates the uplink. Source: WBA(Wireless Broadband Alliance)

Wireless ChannelUplink TrafficDownlink TrafficStatically Assign Higher Priority to Downlink Traffic

And consequently a solution can be to statically assigned higher channel access priority to WiFi AP. Such a solution is not suited for low traffic intensity at the WiFi AP which keeps fluctuating over duration of time.We desire a solution that is adaptive to dynamics of this traffic intensity

Source: WBA(Wireless Broadband Alliance)

Wireless ChannelUplink TrafficDownlink TrafficDynamically Assign Higher Priority to Downlink Traffic

We proposed this solution in which channel access priority for downlink traffic is provided on-Demand. Source: WBA(Wireless Broadband Alliance)Our SolutionSo lets now discuss our WiFox solution in more detailsSource: WBA(Wireless Broadband Alliance)Priority ControlPacket AACKDIFSDIFS HChannel AccessN SlotsN SlotsBusy mediumBusy mediumWins ContentionSTA CSTA BSTA ASmaller IFSClassCWminCWmaxAIFSTXOP LimitAP15164STAs510N/AN/AWe start with understanding how channel access priority can be controlled.In this figure STA C has smaller IFS as compared to STAB and as a result it wins the contention for channel access. We used a combination of 4 such parameters available through 802.11e to ensure that WiFi AP wins the contention for channel access whenever desired. Source: WBA(Wireless Broadband Alliance)Linear Scaling Priority ModelPriority Level

Linear relationship between Goodput and Priority LevelThe WiFi AP operates in two states: Default and HIGH and we define priority level as number of slots on a scale of 10 for which it operates in HIGH state.In this figure we observe the goodput of a node with prioritized channel access for increasing priority level . Source: WBA(Wireless Broadband Alliance)Adaptive Prioritization100 msDHDHDDDHDDTimeHigh PriorityDefault PriorityDecision PointsPriority Level 3As we discussed before the importance of assigning higher priority to downlink traffic in an adaptive manner. WiFox module makes decisions every 100 ms . It observes the TxQ at the WiFi AP to decide the priority level for that decision point. As shown in this example WiFox decides the priority level of 3 and randomly choses 3 slots for HIGH channel access prioritySource: WBA(Wireless Broadband Alliance)EvaluationGive a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)Test Bed2600 Sqft Area with multiple APs, 45 STAsNetgear 802.11 b/g wireless cards with Atheros chipsets and MADWIFI driversLatency emulation using DummyNetModified SURGE for web traffic generation Requests inter-arrival closely follows the ones observed for SIGCOMM tracesUplink UDP traffic using Iperf to emulate Background Traffic

Give a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)Performance

Downlink: N/W GoodputWiFoxW/O WiFoxSignificant improvement in Networks Downlink GoodputGive a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)PerformanceExperiment involves sending 25 requests and observe response for 4 minutes durationRequest Serving rate is 4 times better than NPC

Give a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)Robustness

Performance in presence of Multiple Aps ??WiFoxw/o WiFoxGive a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)Robustness

w/o WiFoxWiFoxUnfair DistributionFairness Realization ??Scenarios where few users indulge in heavy download activitiesTested with one user requesting 150 Mb sized file Source: WBA(Wireless Broadband Alliance)Performance: TxQ Dynamics

WiFoxw/o WiFoxGive a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)ConclusionWiFox Delivers:Deployment Ready SolutionEnhanced user experience with 400-700% Downlink Goodput improvements40-60% faster response timeOpen Problems:Characterizing asymmetry problem for 802.11nSupport for real time applications like chats etc.QoS

30Merci !!Source: WBA(Wireless Broadband Alliance)Multi AP ScenarioDDDDDDDDDDDDDAP 1 ( Priority Level 4)AP 2 ( Priority Level 3)DDDDDDDDDDDDAP 1 ( Priority Level 3)AP 2 ( Priority Level 5)time100 msGive a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)Performance: InsightEnables AP to switch to HIGH priority state under heavy load Avoids TxQ saturationSignificant reduction in ReTx rate compared to stock WiFi implementation

Give a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)Robustness: Uplink TrafficScenarios where few users indulge in heavy uplink activities like video uploading, cloud synchronization etc.

Give a brief overview of the presentation. Describe the major focus of the presentation and why it is important.Introduce each of the major topics.To provide a road map for the audience, you can repeat this Overview slide throughout the presentation, highlighting the particular topic you will discuss next.Source: WBA(Wireless Broadband Alliance)Observations

Source: Rodrig et al.Rodrig et al. analysed SiGCOMM conference network traces to infer increasing retransmission probability with growing number of active clientsSource: WBA(Wireless Broadband Alliance)