(slides) video ads dissemination through wifi-cellular hybrid networks

Video Ads Dissemination throughWiFi-Cellular Hybrid Network (revised)

Hiroshi Hanano1, Yoshihiro Murata2, Naoki Shibata3,Keiichi Yasumoto1 and Minoru Ito1

1Nara Institute of Science and Technology2Hiroshima City University

3Shiga University

Organization

1. Background

2. Related Work

3. Problem Formulation

4. Proposed Algorithm

5. Evaluation through Simulations

6. Conclusion

2Video Ads Dissemination through WiFi-Cellular Hybrid Networks

Demand for Context-Aware Video Ad Dissemination

• Video delivery to mobile terminals has already become common – Video advertisement dissemination now attracting public attention

• Many studies for obtaining user contexts (situation, etc)with sensors, user data (profiles, schedules, etc), and so on

• Context-aware video ad dissemination – Much higher advertisement effect than without context-

awareness

3

Movie trailers

Video Ads Dissemination through WiFi-Cellular Hybrid Networks

Challenges

• Contexts are different among users, change over time– Each video ad should be delivered to users at different

time

4

→ larger data amount transferred via cellular network→ need to consider influence to other cellular services→ number of ads may be restricted

We need a dissemination method whichdoes not overload cellular network

We need a dissemination method whichdoes not overload cellular network


Goals

• Distribute communication load for video ad dissemination from cellular network to WiFi network as much as possible

5

CellularNetwork

load

WiFi(Ad-hoc)


Organization

1. Background

2. Related Work




6. Conclusion


Complementing Cellular Network with MANET

– Some mobile terminals in congested cell area use light loaded cell via MANET

7

Hongyi, W., Chunming, Q., Swades, D. and Ozan, T.:”Integrated Cellular and Ad Hoc Relaying Systems:iCAR,”Proc. of ICC 2001

Hongyi, W., Chunming, Q., Swades, D. and Ozan, T.:”Integrated Cellular and Ad Hoc Relaying Systems:iCAR,”Proc. of ICC 2001

It does not reduce overall data amount transferred for video ads disseminationIt does not reduce overall data amount transferred for video ads dissemination


Introducing BitTorrent into MANET

• Purpose: efficient file exchange among mobile terminals– Realized BitTorrent mechanism on MANET– Evaluation done in terms of mobility and chunk sizes– Show robustness against network partition

8

S. Rajagopalan. and C.-C. Shen:”A Crosslayer Decentralized BitTorrent for Mobile Ad hoc Networks,”Proc. of MOBIQUITOUS 2006

S. Rajagopalan. and C.-C. Shen:”A Crosslayer Decentralized BitTorrent for Mobile Ad hoc Networks,”Proc. of MOBIQUITOUS 2006

It only uses MANET, and thus achievespoor performance on data acquisition ratio/time

It only uses MANET, and thus achievespoor performance on data acquisition ratio/time


Organization

1. Background

2. Related Work




6. Conclusion


Assumptions

• Many users with mobile phones in urban area– Phones can use both cellular and Wi-Fi (ad-hoc mode) networks

• Terminal automatically sets, according to user’s context:– Video ad contents (by file names)– Reception Deadline for each content

10

ex) Context: Bob has a plan to go to a movie after workContents: Trailers of movies currently runningDeadline: Time when work finishes

ex) Context: Bob has a plan to go to a movie after workContents: Trailers of movies currently runningDeadline: Time when work finishes


Target Problem (see PerCom09 paper for more formal description)

• Input– Video ads with deadlines requested by each mobile node

• Output– Actions of mobile nodes at each moment

• Download, Broadcast, or Receive (part of) data

• Constraints– Cellular tower capacity (N connections can be admitted)– Cellular network bandwidth per connection– Wi-Fi radio range of each mobile node– Wi-Fi bandwidth shared by all nodes in each radio range– Reception deadline of each content

• Objective function– Minimize data amount transferred through cellular

network11Video Ads Dissemination through WiFi-Cellular Hybrid Networks

Organization

1. Background

2. Related Work




6. Conclusion


Approach

• File division (like BitTorrent)– Divide each ad content into chunks with fixed data size– Every node can contribute ad dissemination in fair

manner

• Utilization of WiFi (ad-hoc mode)– Reduce load in cellular network by letting nodes exchange chunks through Wi-Fi network

• Fully distributed algorithm– Each terminal decides its action based only on

information of neighbor nodes – Scalable to number of nodes, number of contents, etc


Possible Actions Taken by Mobile Terminal

• Download a chunk through cellular network

• Broadcast a chunk through WiFi (ad-hoc mode) network

• Wait to receive a chunk through WiFi network


Algorithm for Action Decision

• Each node takes an action for global benefit– For each chunk, only one terminal should download it– In each radio range, we should avoid two or more

terminals broadcasting the same chunk

• Flow of proposed algorithm

15

Info Gathering Part

Gather neighboring info

Action Decision Part

Select a chunk

Take an action

Categorize chunks/compute impact of chunks


Information Gathering Part

• Information required to decide action– Demand Info. ： which contents each node requires– Chunk Info. ： which chunks each node retains

• Create Neighbor Table by exchanging hello messages– Each node periodically (ex. 30s) broadcast hello message

including demand information and chunk information

16

a1 a2 a3 …A, B

DemandInfo.

Chunk Info.

:retain :no

Hello Message

a1 a2 a3 …A, Cb1 b2 b3 …B, Da1 a2 a3 …A, D

n1

n2

n3

↑Terminal ID

Neighbor TableVideo Ads Dissemination through WiFi-Cellular Hybrid Networks

Action Decision Part

1. Select a chunk for action– Categorize chunks• Chunks ALREADY obtained broadcast via WiFi• Chunks NOT obtained download from cellular

– Impact computation• Prioritize chunks by putting impact values

– Roulette selection• Select a chunk by roulette

2. Take an action for the selected chunk– For broadcast, we let each node take action at

probability to avoid collision


Action Decision Phase – Categorizing Chunks–• C1: Chunks to be downloaded from cellular– Node wants a chunk, but no neighbor nodes retain it

• C2: Chunks to be broadcast through WiFi– Node retains a chunk, and some of neighbor nodes

want it

• C3: Other chunks– Node does not take any action to those chunks

18

Node underconsideration

Node underconsideration

Neighbor nodes

Neighbor nodes


Action Decision Phase – Impact Computation(1)–• C1: Chunk not retained by any neighbor nodes

• Num. of neighbor nodes that want chunk s:

19

Impact for chunk s:

3

1

3

1 3

1

Avoid redundant download


Action Decision Phase – Impact Computation(2)–• C2: Chunk which some nodes retain, and some of

neighbor nodes want the chunk

• Num of neighbor nodes broadcasting chunk s:

20

Impact for s:

2

1

2

1

Avoid redundant broadcast


Action Decision Phase – Roulette Selection–

• Roulette Selection– Prepare two roulettes for cellular and WiFi– Decide portion of each chunk in proportion to impact– Select a chunk by roulette selection– Probability of selecting a chunk with impact wi

21

High impact chunkLow impact chunk


Problem in Basic Algorithm

• Initially, all nodes intensively use cellular network while

neighboring information is not known• As time elapses, cellular is not used (neighbor nodes have

chunks)

22

Time

100

Time

Not adaptive to remaining time for deadlineNot adaptive to remaining time for deadline

Short increase cellular usage to meet the deadlineLong reduce cellular usage so that nodes exchange chunks via WiFi

DownloadRate (%)

CellularUsage


Adaptation Mechanism (Extended from PerCom09 WiP)

• Consider target download yield at elapsed time to deadline– Current download yield is higher than this, do not use

cellular– Current download yield is lower than this, use cellular

23

100

Elapsed Time deadline

： target DL yield

Not use cellular

Use cellular

Download Yield (%)


Organization

1. Background

2. Related Work




6. Conclusion


Evaluation Experiments

Preliminary Experiment– Measure performance, changing hello message period h and

collision avoidance rate – Purpose: obtain good values for h and

Experiment1– Measure cellular network usage and download time for various

deadlines– Purpose: show adaptability for various deadline lengths

Experiment2– Measure cellular network usage and download time, changing

number of contents – Purpose: show adaptability for various number of ad contents

Metric: cellular network usage ratio (cellular usage)

25

Extensionfrom PerComversion

Cellular usage (%) =amount actually downloaded via cellular

amount when all nodes download all chunks via cellularVideo Ads Dissemination through WiFi-Cellular Hybrid Networks

Common Simulation Parameters

• We developed our own simulator for simulations in 1ms granularity

26

Common simulation parametersSize of each video ad 1.5MB (divided to 1000 chunks)Size of each chunk 1.5KB (sent in a packet)Num of ads that each node requests 2 (selected by Zipf distribution)

Mobility Random Way Point (4km/hour)Field size 500m×500m


Network Model

27

Cellular network Rate/connection: 1.2MbpsMaximum connections: 64

WiFi(ad-hoc mode)

Bandwidth/range: 1.7MbpsRadio range: 200mFading model: Nakagami fadingMAC: CSMA/CA

M. Killat, F. Schmidt-Eisenlohr, H. Hartenstein, C. Rössel, P. Vortish, S. Assenmacher and F. Busch:”Enabling Efficient and Accurate Large-Scale Simulations of VANETs for Vehicular Traffic Management,”Proc. of VANET2007

Packet ReceptionRate (%)


Experiment1: Adaptability to Deadline Lengths

• Evaluate adaptability to various deadline lengths

• Suppose cold start case – All nodes simultaneously start acquiring video ads

after each node decides two ad contents and their deadlines

– Hot start case is explained later

31

Parameters used in Experiment1Number of ad contents 10Number of nodes 500Deadlines 15, 30, 60, 90 minutes

from starting time (fixed)h, 5s, 0.0015


Result of Experiment1 (DL Yield over Time)

• Without deadline adaptation mechanism– DL yield increases slowly after 90% . Many nodes miss deadlines

• With deadline adaptation mechanism– Download yield increases as expected. All nodes meet their deadlines

32

Time

DLYield(%)

Without mechanism

With mechanism(deadline=90min)

60min

30min15min


Result of Experiment1 (Cellular Usage over Time)

• Without deadline adaptation mechanism– Uses cellular network intensively only in initial phase

• With deadline adaptation mechanism– Constantly uses cellular network, but increases usage in

finishing phase• In end phase, it takes long time to receive blank chunks via WiFi

33

40.5%

16.9%

4.3%1.1%

Time

Without mechanism

15min

30min

60min 90min

Number of chunksdownloaded via cellular


Experimental Result with “Hot Start” Configuration

• Mobile nodes join and leave the system• New nodes starts obtaining video ads just after joining• Measured for 267 nodes appearing after 200 minutes elapses

– Cellular usage was reduced further than “cold start” case– Minimum cellular usage was reduced to about 0.3%

34

Deadlines

CellularUsage (%) Cold start

Hot start

ConfigurationInitially 500nodesPoisson arrivals of 15s120min lifetime10 ad contentsDeadlines: 15, 30, 60 (random)


Experiment2: Scalability for Number of Contents

• In actual situation, many video ads may exist– Investigate if our method adaptively works for

different number of ad contents

35

Parameters for Experiment2Number of Ad Contents 10, 20, 30, 40Mobile Terminals 500Time to Deadline 60 minutes


Result of Experiment2

• When increasing number of ad contents– Smaller number of terminals want the same ad– Our method adaptively regulated cellular usage

• All terminals completed downloading all required contents

– Even for 40 contents, cellular usage reduced to about 34.9%36

34.9%

Number of ad contents

Time

Cellular usage (%)

# of chunksvia cellular

40 contents

30

20

10


Conclusion• Cooperative download method for mobile phones for context-

aware video ad dissemination– Does not suppress cellular network by using Wi-Fi network

• Evaluation results supposing typical urban scenarios– Cellular network usage was reduced by up to 98.9%– In hot start scenario, the usage was reduced further– Adaptive to various lengths of deadlines and number of

video ad contents

Future Work– Experiment with realistic situation (real traces, real maps,

etc)– Incentive for mobile terminals– multi-hop paths in ad hoc network


Thanks


Contact :

(slides) video ads dissemination through wifi-cellular hybrid networks

Technology

video ad contents

dissemination method

cellular services number

wifi network

network partition

algorithm evaluation

mobile node wifi

manet evaluation