demonstration of networkcoverage – a mobile network performance measurement app · pdf...

2
Demonstration of NetworkCoverage – A Mobile Network Performance Measurement App Fabian Kaup, Florian Jomrich, David Hausheer Peer-to-Peer Systems Engineering Lab, Technische Universit¨ at Darmstadt Email: {fkaup|fjomrich|hausheer}@ps.tu-darmstadt.de Abstract—Optimizing the Quality of Experience (QoE) of mobile applications over cellular networks requires detailed knowledge of the underlying network and it’s performance. Parameters of interest are, besides the signal strength and availability of technologies, the Round-trip Time (RTT) and available throughput of individual cells at a given location. This information is generally not readily available. Therefore, an Android application measuring the cellular network performance was developed. This demonstration shows the NetworkCoverage App 1 , being implemented to provide visual feedback of the measured network quality to the users, and sample the cellular network in an efficient manner to later support traffic scheduling improvements based on reliable data. I. I NTRODUCTION Improving the energy efficiency and QoE of mobile appli- cations requires detailed knowledge of the cellular network performance. One approach is aggregating delay tolerant network requests to be executed at time periods and locations where the network quality is predicted to be high [1]. Therefore, detailed information on the RTT and throughput of the available networks at the user’s location are required. In the case of a larger measurement set, also the time of the day and periodically repeating traffic patterns are to be considered. Information on the coverage of the individual operator’s networks can usually be found on their website, but lack the required accuracy. Maps of the RTT or network throughput are not readily available. Hence, an Android application, termed NetworkCoverage App, has been developed. There are already a number of applications available monitoring and mapping the mobile network quality. Examples of commercial applications are OpenSignal 2 and Sensorly 3 , measuring the signal strength and to a limited extent also network latency (RTT) and throughput. An academic example is CobCel [2], measuring the signal strength only. NetRadar [3] includes the same metrics, and also creates network maps based on cellular throughput measurements. Contrary, the demonstrated application focuses on reducing the required data volume to measure the network throughput, and accurate localization of the samples, allowing to acquire a larger number of better localized samples. Neidhardt et al. [4] use a crowd-sourcing based approach to derive an accurate positioning algorithm based on observations of given cell IDs. Contrary, the approach demonstrated here measures the performance of each cell and such enables the estimation of the expected network quality in a given cell. 1 https://play.google.com/store/apps/details?id=de.tudarmstadt. networkcoverage [accessed 2014-12-10] 2 http://opensignal.com/android/ [accessed 2014-1-27] 3 http://sensorly.com/ [accessed 2014-10-27] The main novelty of the demonstrated approach is a reduced sampling time, allowing the measurement with higher move- ment velocities. Furthermore, the detailed network information is displayed to the user allowing to determine performance bottlenecks where they appear. The demonstrated application is based on an active network measurement approach, comple- menting passive measurements readily available on the device. Together with these, the time of the measurement, and the exact location for each sample is stored. The measurements are executed on smartphones in different locations and are collected on a central server. There, the col- lected data is stored in raw format and periodically aggregated to analyze the network quality. The aggregated data is made available to mobile devices to predict the performance of a network request at a particular location and time. Based on this data, improvements of the mobile energy consumption and QoE as described in [1] are possible. The remainder of this paper gives additional details on the NetworkCoverage App in Section II. The server based data aggregation is described in Section III. Section IV explains the demo setup and Section V concludes the paper. II. NETWORKCOVERAGE APP The NetworkCoverage App, a closed source, but freely available App, was developed with focus on crowd-sourcing data collection. It includes a number of features visualizing the network performance and the ongoing measurements. These include a map of the collected data, showing signal strength, encountered network technologies (i.e. GPRS, UMTS, LTE, ...) and in the case of active measurements the RTT to the closest server of the EmanicsLab 4 testbed, the TCP downlink throughput, or both. Furthermore, this information is displayed on a status screen, and using a widget, which can be placed on the home screen. Figure 1 shows two screenshots of the Android application. Figure 1a shows a map with an overlay of the aggregated throughput measurements as returned by the server. Further- more, the locally collected data is displayed on the map, allowing the detailed insight into the network performance for a given situation. Figure 1b shows the widget, letting the user know about the current network quality and execute throughput measurements from his home screen. The signal strength measurements are executed automati- cally, monitoring the network quality without user interaction. The RTT and throughput measurements are generally executed 4 http://www.emanicslab.org/ [accessed 2014-10-31]

Upload: ngongoc

Post on 21-Mar-2018

217 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Demonstration of NetworkCoverage – A Mobile Network Performance Measurement App · PDF file · 2018-01-30Demonstration of NetworkCoverage – A Mobile Network Performance Measurement

Demonstration of NetworkCoverage – A MobileNetwork Performance Measurement App

Fabian Kaup, Florian Jomrich, David HausheerPeer-to-Peer Systems Engineering Lab, Technische Universitat Darmstadt

Email: {fkaup|fjomrich|hausheer}@ps.tu-darmstadt.de

Abstract—Optimizing the Quality of Experience (QoE) ofmobile applications over cellular networks requires detailedknowledge of the underlying network and it’s performance.Parameters of interest are, besides the signal strength andavailability of technologies, the Round-trip Time (RTT) andavailable throughput of individual cells at a given location. Thisinformation is generally not readily available. Therefore, anAndroid application measuring the cellular network performancewas developed. This demonstration shows the NetworkCoverageApp1, being implemented to provide visual feedback of themeasured network quality to the users, and sample the cellularnetwork in an efficient manner to later support traffic schedulingimprovements based on reliable data.

I. INTRODUCTION

Improving the energy efficiency and QoE of mobile appli-cations requires detailed knowledge of the cellular networkperformance. One approach is aggregating delay tolerantnetwork requests to be executed at time periods and locationswhere the network quality is predicted to be high [1]. Therefore,detailed information on the RTT and throughput of the availablenetworks at the user’s location are required. In the case of alarger measurement set, also the time of the day and periodicallyrepeating traffic patterns are to be considered.

Information on the coverage of the individual operator’snetworks can usually be found on their website, but lack therequired accuracy. Maps of the RTT or network throughput arenot readily available. Hence, an Android application, termedNetworkCoverage App, has been developed. There are alreadya number of applications available monitoring and mapping themobile network quality. Examples of commercial applicationsare OpenSignal2 and Sensorly3, measuring the signal strengthand to a limited extent also network latency (RTT) andthroughput. An academic example is CobCel [2], measuring thesignal strength only. NetRadar [3] includes the same metrics,and also creates network maps based on cellular throughputmeasurements. Contrary, the demonstrated application focuseson reducing the required data volume to measure the networkthroughput, and accurate localization of the samples, allowing toacquire a larger number of better localized samples. Neidhardtet al. [4] use a crowd-sourcing based approach to derive anaccurate positioning algorithm based on observations of givencell IDs. Contrary, the approach demonstrated here measuresthe performance of each cell and such enables the estimationof the expected network quality in a given cell.

1https://play.google.com/store/apps/details?id=de.tudarmstadt.networkcoverage [accessed 2014-12-10]

2http://opensignal.com/android/ [accessed 2014-1-27]3http://sensorly.com/ [accessed 2014-10-27]

The main novelty of the demonstrated approach is a reducedsampling time, allowing the measurement with higher move-ment velocities. Furthermore, the detailed network informationis displayed to the user allowing to determine performancebottlenecks where they appear. The demonstrated applicationis based on an active network measurement approach, comple-menting passive measurements readily available on the device.Together with these, the time of the measurement, and the exactlocation for each sample is stored.

The measurements are executed on smartphones in differentlocations and are collected on a central server. There, the col-lected data is stored in raw format and periodically aggregatedto analyze the network quality. The aggregated data is madeavailable to mobile devices to predict the performance of anetwork request at a particular location and time. Based onthis data, improvements of the mobile energy consumption andQoE as described in [1] are possible.

The remainder of this paper gives additional details on theNetworkCoverage App in Section II. The server based dataaggregation is described in Section III. Section IV explains thedemo setup and Section V concludes the paper.

II. NETWORKCOVERAGE APP

The NetworkCoverage App, a closed source, but freelyavailable App, was developed with focus on crowd-sourcingdata collection. It includes a number of features visualizing thenetwork performance and the ongoing measurements. Theseinclude a map of the collected data, showing signal strength,encountered network technologies (i.e. GPRS, UMTS, LTE,...) and in the case of active measurements the RTT to theclosest server of the EmanicsLab4 testbed, the TCP downlinkthroughput, or both. Furthermore, this information is displayedon a status screen, and using a widget, which can be placedon the home screen.

Figure 1 shows two screenshots of the Android application.Figure 1a shows a map with an overlay of the aggregatedthroughput measurements as returned by the server. Further-more, the locally collected data is displayed on the map,allowing the detailed insight into the network performance fora given situation. Figure 1b shows the widget, letting the userknow about the current network quality and execute throughputmeasurements from his home screen.

The signal strength measurements are executed automati-cally, monitoring the network quality without user interaction.The RTT and throughput measurements are generally executed

4http://www.emanicslab.org/ [accessed 2014-10-31]

Page 2: Demonstration of NetworkCoverage – A Mobile Network Performance Measurement App · PDF file · 2018-01-30Demonstration of NetworkCoverage – A Mobile Network Performance Measurement

(a) Throughput map and details of alocal measurement

(b) Widget showing the current net-work information

Fig. 1. Screenshots of the NetworkCoverage App

manually, but can also be configured to run periodically. Thisis necessary, as these may consume a high amount of datavolume, depending on the available network technologies, anddata plans are often limited. A background service is used toallow measurements independent of the user activity. It handlesindividual tasks like location monitoring, network informationmonitoring, and data storage in the local DB. Furthermore,the automatic RTT and throughput measurements and dataupload to the server are triggered. Incoming measurementsare sent to the user activity using broadcasts. To reduce theinfluence of the data upload on the mobile data cap, the usermay choose to send measurement samples on WiFi only. Onlyafter a successful transmission, the respective table columnsare marked as sent, ensuring that no data is lost.

III. SERVER-BASED DATA AGGREGATION

The collected data is stored on a central server. This allowsaggregating measurements from multiple mobile devices, andderive further network metrics like cell densities and sizes.

The aggregated network information is available to the end-user in the form of map tiles. These are generated for thesignal strength, RTT, and throughput for each network operatorindividually. Here, the difference between network provider andnetwork operator is important, as multiple network providersmight offer services based on the network run by one operator.This allows also aggregating measurements of different re-sellers. These maps are available within the mobile applicationand on a website. Knowing the network quality, mobile userscan adapt their data consumption to improve the networkperformance and save energy on the mobile device. Furthermore,participating users may decide to execute measurements atpreviously unsurveyed locations, helping others to get anoverview of the network quality.

Aggregating a large number of individual measurements,the size and number of cells covering a given location can bederived as well. From the number of cells, the overall available

bandwidth at a given location can be derived. This can also beincluded in the mobile bandwidth prediction, as the probabilityof congestion within multiple cells is lower compared to asingle cell. Also from the cell sizes at the current locationconclusions on the network availability can be drawn, as forsmall cells the expected throughput is higher due to a lowernumber of devices in a single cell.

IV. DEMO SETUP

The proposed demo will show the NetworkCoverage Appon the smartphone. This includes the network status view,visualizing the signal strength and available technologies,the measurement procedure for throughput measurements,showing the saved traffic compared to conventional approaches,and the map view containing the server generated signalstrength, throughput and RTT maps as well as the localmeasurements with detailed information on the measurements.Live measurements will be conducted and the functionality ofthe improved measurement procedure displayed and comparedto conventional throughput measurements using iPerf.

Furthermore, the server front-end will be demonstrated ona laptop. The maps available there will show the same maps asthe Android application, and will let the demo visitors discoverdifferent aspects of the network quality, including the cellcoverage and densities. A good demo experience is achievedby executing measurements in the vicinity of the conferencevenue before the event.

V. SUMMARY

The demonstrated Android application allows the measure-ment of the cellular and WiFi network quality by executingactive and passive measurements. The improvements of mea-surement speed and saved data volume are visualized andcompared to conventional approaches. The collected data isstored locally for in situ network analysis, and is periodicallytransmitted to a server for aggregation and global analysis ofthe network performance. The aggregated data, as fetched fromthe server, is displayed on the local device in the form ofnetwork quality maps, including RTT and throughput maps ofthe different network operators. From this, the end-user maydeduce the available network quality to improve the experiencednetwork quality and battery lifetime of their devices.

ACKNOWLEDGMENTS

This work has been supported in parts by the EU(FP7/#317846, SmartenIT and FP7/#318398, eCOUSIN) andthe DFG as part of the CRC 1053 MAKI. The authors wouldlike to acknowledge valuable comments by their colleaguesand project partners.

REFERENCES

[1] F. Kaup and D. Hausheer, “Optimizing Energy Consumption and QoEon Mobile Devices,” in IEEE ICNP, 2013.

[2] J. Pino and J. E. Pezoa, “CobCel: Distributed and Collaborative Sensingof Cellular Phone Coverage Using Google Android,” in ICSNC, 2012.

[3] S. Sonntag, L. Schulte, and J. Manner, “Mobile Network Measurements –It’s not all about Signal Strength,” in IEEE WCNC, 2013.

[4] E. Neidhardt, A. Uzun, U. Bareth, and A. Kupper, “Estimating Locationsand Coverage Areas of Mobile Network Cells Based on CrowdsourcedData,” in IEEE IFIP WMNC, 2013.