modular wireless networks for infrastructurally -challenged environments

Modular Wireless Networks for Infrastructurally-Challenged

EnvironmentsPhD Proposal TalkMariya Zheleva

PhD Advisor:

Committee Members:

Professor Elizabeth Belding, UCSB

Professor Heather Zheng, UCSBProfessor Ben Zhao, UCSBRanveer Chandra, Microsoft Research

2Connectivity as an Enabler

Economic growth High-quality workforce Improved job-search Purchasing ability

[internetinnovation.org][Silicon Alley Insider]

Education Online education Remote tutoring

Healthcare Disease spread prediction

3Affordability and Accessibility

Source: ITU

Prepaid handset-based (500MB)

Postpaid handset-based (500MB)

Prepaid computer-based (1GB)

Postpaid computer-based (1GB)

0 10 20 30 40 50 60 701.3

1.2

2.2

1.4

15.7

11.3

24.7

18.8

11.4

7.5

17.7

12.7

38.8

36.2

58.3

54.6

Price of mobile broadband services, 2013(% from monthly income)

Africa World Developing Developed

An African may pay 39 times more than a Westerner for Internet access.

4Affordability and Accessibility

Source: ITU

5Subscription Trends

Mobile subscriptions approach the world population

Plain Cellular

Developing world subscription stats do not capture the number of people with access

People carry multiple SIM cards Pricing Coverage availability

6Subscription Trends

Computer-basedMobile-based

BroadbandBroadband access is still very limited.

7

“Mobile network operators find it commercially infeasible to operate in rural areas.”

World Bank’s 2012 report on maximizing mobile

8Why Commercially-Infeasible?

9What if the call could be local?

10Thesis Statement

In order to optimize user experience, rural networks and applications need to be highly modular and use this modularity to switch between global and local context according to network conditions and demand.

In order to optimize user experience, rural networks and applications need to be highly modular and use this modularity to switch between global and local context according to network conditions and demand.

In order to optimize user experience, rural networks and applications need to be highly modular and use this modularity to switch between global and local context according to network conditions and demand.

11

Identifyingchallenges

Connectivitysolutions

Sociological Approach(interviews)

Cellular network traces analysis

Internet traffic analysis

Enabling network performance analysis

Bringing visibility to rural users [ICTD13]

Community detection in cellular [ICTD13]

Internet usage [IJoC12]

Cellphone usage [MobiSys13]

Cellphones in healthcare [GSWC12]

The bandwidth divide [IJoC12]

The increased bandwidth fallacy

[ACM DEV13] and WiP

AirLab [SIGCOMM CCR11]

Technological Approach

Connectivity

Long distance Last mile

PHY/MAC Layer Application Layer

Applications

VillageLink [COMSNETS14]

HybridCellWiP

Kwiizya [MobiSys13]

ImmuNet [GSWC12]

*WiP – Work in Progress

Completed

In progress

TxMinerWiP

Bringing visibility to rural users [ICTD13]

The increased bandwidth fallacy

[ACM DEV13] and WiP

Kwiizya [MobiSys13]

Identifying Challenges

13Methodology: Sociological Approach

Goal:• Understand cellphone usage• Cellphones in healthcare

Methodology:• In-person interviews• Volunteer at the hospital in Macha

14Methodology: Technological Approach

Goal:• Cellphones (rural vs. urban) :

• Usage• Provisioning• Community detection

• Internet:• Usage• Performance• Avenues for improvement

Methodology:• Cellular: Analysis of large-scale datasets• Internet: Rural network traffic analysis

Antenna-to-antenna statistics

Egocentric graphs over

time

Identifying Challenges: Internet Usage

16Network ArchitecturePre-upgrade

Macha WLAN Internet

Switch

MonitoringServer

Gateway

Post-upgrade

Macha WLAN Internet

Switch

MonitoringServer

Gateway

DL: 256kbps (bursting to 1Mbps)

UL: 64kbps

2Mbps

[Zheleva, Schmitt, Vigil, Belding, ACM DEV 2013]

17

PostPre

Background of Analysis

April 9th, 2011

Long-Term (LT)

March 9th – April 9th, 2011

April 9th – May 9th, 2011 July, 2011

While bandwidth increase may improve network usability, it is not proportional to improved user experience.

[Zheleva, Schmitt, Vigil, Belding, ACM DEV 2013]

time

18Implications on Content Access/Generation

TCP Flow analysis: Success and Failure of TCP Uploads and Downloads

[Zheleva, Schmitt, Vigil, Belding, ACM DEV 2013]

x4

19Usage: Top Web Sites

[Zheleva, Schmitt, Vigil, Belding, ACM DEV 2013]

20Usage: Top-4 Services Success and Failure

[Zheleva, Schmitt, Vigil, Belding, ACM DEV 2013]

*200 – OK 400 – Bad Request 408 – Request Timeout

21The Increased Bandwidth Fallacy

• Increased bandwidth can improve usability• Bandwidth increase not proportional to improved user experience• Prioritize bandwidth for critical services• E.g. software updates

[Zheleva, Schmitt, Vigil, Belding, ACM DEV 2013]

Internet Local WLAN

Increased 400 and 408 + Failure to update Mozilla

= Prioritize SW update

22

Identifyingchallenges

Connectivitysolutions

Sociological Approach(interviews)

Cellular network traces analysis

Internet traffic analysis

Enabling network performance analysis

Bringing visibility to rural users [ICTD13]

Community detection in cellular [ICTD13]

Internet usage [IJoC12]

Cellphone usage [MobiSys13]

Cellphones in healthcare [GSWC12]

The bandwidth divide [IJoC12]

The increased bandwidth fallacy

[ACM DEV13] and WiP

AirLab [SIGCOMM CCR11]

Technological Approach

Connectivity

Long distance Last mile

PHY/MAC Layer Application Layer

Applications

VillageLink [COMSNETS14]

HybridCellWiP

Kwiizya [MobiSys13]

ImmuNet [GSWC12]

*WiP – Work in Progress

Completed

In progress

TxMinerWiP

Bringing visibility to rural users [ICTD13]

The increased bandwidth fallacy

[ACM DEV13] and WiP

Kwiizya [MobiSys13]

Identifying Challenges:Bringing Visibility to Rural Cellular Users

24Bringing Visibility to Rural Users

December 1, 2011 April 1, 2012

Aggregate # calls/hourAggregate duration of calls/hour

Antenna IDLatitude

Longitude

Antenna IDLatitude

Longitude

+ AfriPop population density dataset

+ OECD typology

< 150/km2 – Rural150 – 300/km2 – Suburban

> 300/km2 – Urban

[Zheleva, Schmitt, Vigil, Belding, ICTD 2013]

Orange in Cote D’IvoireAntenna-to-Antenna communication

25Bringing Visibility to Rural Users

98% Rural territory2% Urban territory

[Zheleva, Schmitt, Vigil, Belding, ICTD 2013]

26Bringing Visibility to Rural Users

[Zheleva, Schmitt, Vigil, Belding, ICTD 2013]

Urban Rural

23

70

Same antenna calls - w/o TRANSPORTATION

% o

f cal

ls

High locality of rural cellular communications.

51.7% of rural antennas are TRANSPORTATION.

27Bringing Visibility to Rural Users

Mobile cellular services are expensive.Prepaid is prevalent.

Urban Rural23

70

% o

f cal

ls

Rural users are under-provisioned.

High locality of interest.

An African may pay 39 times more than a Westerner for Internet access.

Improve the situation by providing alternatives.

Connectivity Solutions: Local Cellular Network

Services in Remote Areas

29Kwiizya*: Architecture

* Kwiizya means “to chat” in Tonga

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

30Kwiizya: Architecture

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

31Kwiizya: Call Switching

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

32Kwiizya: SMS Switching

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

33Support of SMS Applications (IM-SMS)

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

34Macha, Zambia

Source: GoogleMaps

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

35The Base Station

A RangeNetworks SNAP unit

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

36Installation Sites

LITAWater Tower

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

37Power Quality in Macha

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

38The Power Supply

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

39The Network in Macha

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

40Evaluation• 20 SIM cards• Local partners (IT staff)• Restricted registration• Voice and SMS

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

41Controlled Experiments Setup

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

42Performance: SMS Delay

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

SMS delivery in Tanzania takes between 6 and 16 seconds depending on provider.

43Performance: Call Quality

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

GSM FR (6.10)Min delay 20ms

GSM FR (6.10)Max MOS 3.46

44Performance: IM to SMS

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

45Performance: IM to SMS

[Zheleva, Paul, Johnson, Belding, MobiSys 2013]

46

Identifyingchallenges

Connectivitysolutions

Sociological Approach(interviews)

Cellular network traces analysis

Internet traffic analysis

Enabling network performance analysis

Bringing visibility to rural users [ICTD13]

Community detection in cellular [ICTD13]

Internet usage [IJoC12]

Cellphone usage [MobiSys13]

Cellphones in healthcare [GSWC12]

The bandwidth divide [IJoC12]

The increased bandwidth fallacy

[ACM DEV13] and WiP

AirLab [SIGCOMM CCR11]

Technological Approach

Connectivity

Long distance Last mile

PHY/MAC Layer Application Layer

Applications

VillageLink [COMSNETS14]

HybridCellWiP

Kwiizya [MobiSys13]

ImmuNet [GSWC12]

*WiP – Work in Progress

Completed

In progress

TxMinerWiP

Bringing visibility to rural users [ICTD13]

The increased bandwidth fallacy

[ACM DEV13] and WiP

Kwiizya [MobiSys13]

TxMinerWiP

HybridCellWiP

Proposed Work: HybridCell: Coexistence

of Cellular Networks

48HybridCell: Architecture

L-Cell

C-Cell

Village

*C-Cell – Commercial cell*L-Cell – Local cell

RTL-SDRsniffer

Observe

Control

49

RF Frontend

HybridCell vs. Commercial Network

In-Village Out-Village

Users

Loca

l (Kw

iizya

)Co

mm

erci

al

BTS

BTS

RNC

MSC

SGSN

HLRVLR

GGSN

PSTN

Internet

Core Network

MSC+VLR+HLR + GSN

Internet Gateway

Wireless

Expensive: Highly centralized Hardware Software Backbone

Low-cost: Local SDR hardware Open-source software Generic IP backbone

50HybridCell Use-CasesLocal Services for End Users Local Services for Organizations

Low-Cost Packet-Switched Data Disaster Communications

[Source: BBC]

51HybridCell: Research Challenges

• Understanding the state of rural commercial cellular access:• Measuring signal quality – capture SNR using an RTL-SDR setup• Measuring packet-switched data quality

• HybridCell use with existing SIM.• User-triggered• System-triggered

• Monitoring C-Cell’s activity:• RTL-SDR setup• Use TxMiner

• Mobile data offload through local cells.

Proposed Work: TxMiner: DSA beyond TV

White Spaces

53TxMiner: DSA Beyond TV White Spaces

512MHz 700MHz

-130

dBm

-60d

Bm

Power Spectral Density: TVWS

-130

dBm

-30d

Bm

30MHz 6000MHz

-110dBm

Power Spectral Density: Beyond TVWS

54TxMiner Goal

Power Spectral Density Graph

PSD,

dBm

/Hz

Frequency, MHz

Transmissions:• Bandwidth• TDMA/FDMA• Mobility• Direction

55

• Enhance existing DSA systems

• Support spectrum regulations in developing nations and worldwide

TxMiner Applications

• TX periodicity• TX bandwidth• Mobile TX over time• Primary or Secondary

TxMiner-enhancedDSA Database

Secondary User Network

1) Spectrum availability?

2) Spectrum availability +

Transmitter Characteristics

3) Bandwidth X satisfies user

demand

Geo-location database

56In Conclusion

• Identified challenges associated with access• Cellular networks• Internet

• Designed solutions for local cellular access• With support for SMS applications

• Propose extensions• Cellular networks coexistence• DSA beyond TV White Spaces

Thesis Statement: In order to optimize user experience, rural networks and applications need to be highly modular and use this modularity to switch

between global and local context according to network conditions and demand.

57PhD Timeline NOV DEC JAN FEB MAR APR MAY JUN

Africa conferences and fieldwork

Preparing job applications

TxMiner evaluation and paper submission

HybridCell research

HybridCell paper

Increased bandwidth fallacy, journal PhD dissertation and

defence

Preparing job talk/interviewing

58

Thank you!Mariya Zheleva

mariya@cs.ucsb.edu

This work was done in collaboration with: Elizabeth Belding; Veljko Pejovic; David

Johnson; Arghyadip Paul; Morgan Vigil; Paul Schmitt; Partners from LinkNet, Macha; Kate

Milosavljevic