Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
DTN-based<br />Delivery of Word-of-Mouth Information<br />with Priority and Deadline<br />Yasuhiro Ishimaru, Weihua Sun, <...
Users want to share data even in area where communication infrastructure is not available. <br />(Ex. disaster area, rural...
 By Delay Tolerant Network</li></ul>Target spot:<br />data exist<br />Source spot:<br /> sending request<br />Request<br /...
Limitation in DTN environments<br />Data amount that can be transferred through DTN is limited<br />User may not receive a...
Background<br />Related Work<br />Proposed Method<br />Experiment<br />Conclusion<br />Outline<br />2010/4/26<br />4<br />...
Data delivery based on probabilistic flooding<br />Each node replicates data to encountered nodes with a certain probabili...
Small server with storage & communication functions<br />Deployed at multiple different spots<br />Increase opportunities ...
Existing studies<br />Objective: Improving data delivery ratio, reducing delay<br />Problem: When congestion occurs , deli...
Background<br />Related Work<br />Proposed Method<br />Experiment<br />Conclusion<br />Outline<br />2010/4/26<br />8<br />...
Maximize overall user satisfactionin congested DTN environments<br />Deploy InfoBoxes into target area <br />to increase c...
Sharing word-of-mouth information in rural sightseeing area<br />Put InfoBox at each sightseeing spot<br />Users (tourists...
User enjoys favorite tour by visiting spots and staying there<br />Sends requests and receives reply data through InfoBox<...
Equipped with mobile terminals (cell phones)<br />Capable of Bluetooth communication<br />Active Behavior<br />Moving betw...
Small battery-driven PC equipped with:<br />Sufficient CPU power and storage<br />Bluetooth communication capability<br />...
Communication model<br />Bluetooth-based communication<br />Radio range: circle with radius R (e.g., 10m)<br />No packet l...
InfoBox schedules send/receive actions by applying the following techniques to each data in its queue<br />Delivery time e...
Sending a data which CANNOTarrive by deadline wastes resource, and disturbs other data’s delivery<br />Why estimate delive...
Delivery time estimationfor each data<br />Target Spot<br />Receiving Spot<br />Source Spot<br />staying time ε<br />trave...
To increase the data delivery ratio<br />InfoBox must replicate data to multiple users <br />Too much replication results ...
Based on user’s moving probability</li></ul>Why estimate number of data replicas?<br />Data<br />Data<br />Target InfoBox<...
Decide appropriate number of replicas<br />Route in time <br />for Deadline<br />System Parameter: Required delivery ratio...
Each InfoBox calculates data's delivery efficiency (ECP)<br />ECP denotes importance score per KB<br />Cost-performance es...
Multi Hop</li></ul>ECP<br />  =(50 x 90%) / (100 x 3)<br />  =0.15 (points/KB)<br />2010/4/26<br />20<br />ICMU2010@Seattl...
InfoBox sorts data in descending order of ECP<br />Replicates data to users with calculated no. of replicas<br />Data sche...
Background<br />Related Work<br />Proposed Method<br />Experiment<br />Conclusion<br />Outline<br />2010/4/26<br />22<br /...
We developed own  simulator in Java<br />Simulation Configuration<br />Simulation parameters<br />User mobility model<br /...
Total Satisfaction<br />Overall importance scores associate with data that were delivered within deadline.<br />A larger v...
Total Satisfactionvs. Required delivery ratio(δ)<br />Result:<br /><ul><li>A higherδ leads a higher total satisfaction
Proposed method is delivery C-P sensitive
Achieved significant improvement</li></ul>δ<br />2010/4/26<br />25<br />ICMU2010@Seattle<br />
Upcoming SlideShare
Loading in …5
×
Upcoming SlideShare
地下街におけるスマートフォンの光を用いた避難誘導方式の提案
Next

0

Share

2010-04-24-DTN-based Delivery of Word-of-Mouth Information with Priority and Deadline

2010/04/05 ICMU@Seattle

Related Books

Free with a 30 day trial from Scribd

See all
  • Be the first to like this

2010-04-24-DTN-based Delivery of Word-of-Mouth Information with Priority and Deadline

  1. 1. DTN-based<br />Delivery of Word-of-Mouth Information<br />with Priority and Deadline<br />Yasuhiro Ishimaru, Weihua Sun, <br />Keiichi Yasumoto, Minoru Ito<br />2010/4/26<br />1<br />ICMU2010@Seattle<br />Nara Institute of Science and Technology, Japan<br />
  2. 2. Users want to share data even in area where communication infrastructure is not available. <br />(Ex. disaster area, rural area, etc)<br />Target application<br />data retrieval by specifying target spot and receiving spot<br />Data Retrieval in DTN Environments<br /><ul><li> We target Word-of-mouth information sharing among users in a rural sightseeing area
  3. 3. By Delay Tolerant Network</li></ul>Target spot:<br />data exist<br />Source spot:<br /> sending request<br />Request<br />Request<br />Data is transferred by persons<br />with computing devices<br />Reply<br />Receiving spot:<br />receiving reply<br />2010/4/26<br />2<br />ICMU2010@Seattle<br />
  4. 4. Limitation in DTN environments<br />Data amount that can be transferred through DTN is limited<br />User may not receive all reply data<br />User wants to receive<br />Reply data by deadline(e.g.,event info, time sale info)<br />More important datawhen sending multiple requests<br />Requirements for data sharing in DTN<br />2010/4/26<br />3<br />ICMU2010@Seattle<br />We need a differentiation mechanism that transfers more important/deadline-sensitive data prior to others<br />
  5. 5. Background<br />Related Work<br />Proposed Method<br />Experiment<br />Conclusion<br />Outline<br />2010/4/26<br />4<br />ICMU2010@Seattle<br />
  6. 6. Data delivery based on probabilistic flooding<br />Each node replicates data to encountered nodes with a certain probability and repeats this to disseminate the data <br />Objective is improvement of data delivery ratio<br />Epidemic Routing[3]<br /><ul><li>Improve data delivery ratio by adjusting replication probability</li></ul>Destination node<br />Too much replicas<br />Source node<br />2010/4/26<br />5<br />ICMU2010@Seattle<br />Congestion occurs<br />Delivery ratio down<br />
  7. 7. Small server with storage & communication functions<br />Deployed at multiple different spots<br />Increase opportunities for mobile nodes to exchange data<br />Increase data delivery ratio<br />Throwbox[7]<br />With Throwbox<br />Without Throwbox<br />2010/4/26<br />6<br />ICMU2010@Seattle<br />
  8. 8. Existing studies<br />Objective: Improving data delivery ratio, reducing delay<br />Problem: When congestion occurs , delivery ratio of all data are reduced uniformly<br />Our contribution<br />DTN data delivery method considering deadline and priority<br />Schedule data transfer based on deadline<br />Transfer more important data prior to others<br /><ul><li>Objective: Increase overall user satisfaction</li></ul>Contribution of Our Research<br />2010/4/26<br />7<br />ICMU2010@Seattle<br />
  9. 9. Background<br />Related Work<br />Proposed Method<br />Experiment<br />Conclusion<br />Outline<br />2010/4/26<br />8<br />ICMU2010@Seattle<br />
  10. 10. Maximize overall user satisfactionin congested DTN environments<br />Deploy InfoBoxes into target area <br />to increase communication opportunities<br />By scheduling at InfoBox<br /> Save DTN resource<br />discard data that are likely to miss deadline<br /> Increase user satisfaction<br />transfer the data with higher cost-performance prior to others<br />Goal and Basic Ideas<br />2010/4/26<br />9<br />ICMU2010@Seattle<br />Goal<br />Basic ideas<br />Similar to <br />Throwbox<br />InfoBox<br />
  11. 11. Sharing word-of-mouth information in rural sightseeing area<br />Put InfoBox at each sightseeing spot<br />Users (tourists) retrieve word-of-mouth info via InfoBox<br />Target Environment<br />Todaiji temple<br />Target area<br />Nara Park<br />Nara Station<br />Kasuga Shrine<br />Ukimido Temple<br />User<br />InfoBox<br />2010/4/26<br />10<br />ICMU2010@Seattle<br />
  12. 12. User enjoys favorite tour by visiting spots and staying there<br />Sends requests and receives reply data through InfoBox<br />Request contains: destination spot, receiving spot, importance score<br />User gets satisfaction if it receives reply data at receiving spot<br />Service Model<br />Todaiji<br />Nara Park<br />Rep:20<br />Req:20<br />Nara St.<br />Example request:<br />DestSpot: Nara Park<br />RecSpot: KasugaShrineImportance Score: 20<br />KasugaShrine<br />Ukimido<br />Satisfaction:20<br />2010/4/26<br />11<br />ICMU2010@Seattle<br />
  13. 13. Equipped with mobile terminals (cell phones)<br />Capable of Bluetooth communication<br />Active Behavior<br />Moving between spots and staying at a spot<br />Sending request to InfoBox<br />Passive Behavior<br />Receiving reply data from InfoBox<br />Relaying data between InfoBoxes (receive, carry, re-send)<br />Assumption for User<br />2010/4/26<br />12<br />ICMU2010@Seattle<br />InfoBox<br />
  14. 14. Small battery-driven PC equipped with:<br />Sufficient CPU power and storage<br />Bluetooth communication capability<br />Deployed near gate to each spot<br />CANNOT communicate with other InfoBox<br />Schedules send/receive actions with user terminals<br />Knows user’s moving probability between spots<br />Assumption for InfoBox<br />Sightseeing spot<br />gate<br />InfoBox<br />50%<br />70%<br />InfoBox C<br />50%<br />InfoBoxA<br />80%<br />20%<br />30%<br />InfoBox B<br />13<br />ICMU2010@Seattle<br />
  15. 15. Communication model<br />Bluetooth-based communication<br />Radio range: circle with radius R (e.g., 10m)<br />No packet loss due to collision<br />Max. Available Bandwidth: BW (e.g., 1Mbps) <br />Queue-based communication<br />InfoBox has a queue for storing and sending data<br />Congestion: receive-data amounts > send-able-amounts<br />Assumption for Communication between InfoBox and User<br />A<br />2010/4/26<br />14<br />ICMU2010@Seattle<br />Congestion<br />Receiving<br />Sending<br />Queue<br />B<br />C<br />D<br />
  16. 16. InfoBox schedules send/receive actions by applying the following techniques to each data in its queue<br />Delivery time estimation for the data<br />Decision of appropriate number of replicas for the data<br />Cost-performance estimation for the data<br />These techniques achieve data delivery with high overall user satisfaction<br />Proposed Scheduling Algorithm<br />2010/4/26<br />15<br />ICMU2010@Seattle<br />
  17. 17. Sending a data which CANNOTarrive by deadline wastes resource, and disturbs other data’s delivery<br />Why estimate delivery time?<br />Replication time = 10 min<br />Delivery time = 20 min<br />Replicationtime = 10 min<br /> By using delivery time estimation, we can discard data seems to miss deadline <br />Arrival<br />Delivery time = 20 min<br />Arrival<br />DL: 10 min<br />30min<br />30min<br />DL: 40 min<br />40min<br />30min<br />DL: 10 min<br />DL: 40 min<br />50min<br />40min<br />DL: 40 min<br />InfoBox<br />DL: 80 min<br />60min<br />50min<br />DL: 40 min<br />DL: 80 min<br />Number of delivered data: 2<br />Number of delivered data: 3<br />2010/4/26<br />16<br />ICMU2010@Seattle<br />
  18. 18. Delivery time estimationfor each data<br />Target Spot<br />Receiving Spot<br />Source Spot<br />staying time ε<br />traveling time γ<br />Reply delivery timeβ<br />Request delivery time α<br />ReceivingDeadline = γ+ε<br />Delivery Time = α+β ≦ Receiving Deadline<br />Data which can not satisfy the constraint will be discarded<br />2010/4/26<br />17<br />ICMU2010@Seattle<br />
  19. 19. To increase the data delivery ratio<br />InfoBox must replicate data to multiple users <br />Too much replication results in waste bandwidth<br /><ul><li>Calculates appropriate number of replicas
  20. 20. Based on user’s moving probability</li></ul>Why estimate number of data replicas?<br />Data<br />Data<br />Target InfoBox<br />Replica<br />Replica<br />Source InfoBox<br />30%<br />Replica<br />Replica<br />Replica<br />Replica<br />70%<br />2010/4/26<br />18<br />ICMU2010@Seattle<br />
  21. 21. Decide appropriate number of replicas<br />Route in time <br />for Deadline<br />System Parameter: Required delivery ratio δ<br />Required ratio that data is delivered from Source to Target<br />Based on users moving probability, to achieve δ<br />Ex. δ = 0.9, p1=0.3, p2=0.7<br />Target<br />p1(30%)<br />Route expires <br />of Deadline<br />Source<br />p2(70%)<br />n(=7) is the appropriate number of replicas<br />2010/4/26<br />19<br />ICMU2010@Seattle<br />
  22. 22. Each InfoBox calculates data's delivery efficiency (ECP)<br />ECP denotes importance score per KB<br />Cost-performance estimation<br /><ul><li>Single Hop
  23. 23. Multi Hop</li></ul>ECP<br /> =(50 x 90%) / (100 x 3)<br /> =0.15 (points/KB)<br />2010/4/26<br />20<br />ICMU2010@Seattle<br />
  24. 24. InfoBox sorts data in descending order of ECP<br />Replicates data to users with calculated no. of replicas<br />Data scheduling by proposed method<br />A<br />A<br />ECP=0.8<br />Replica=3<br />B<br />ECP=0.7<br />Replica=4<br />C<br />ECP=0.5<br />Replica=4<br />A<br />Sending Queue<br />A<br />Other Spots<br />2010/4/26<br />21<br />ICMU2010@Seattle<br />
  25. 25. Background<br />Related Work<br />Proposed Method<br />Experiment<br />Conclusion<br />Outline<br />2010/4/26<br />22<br />ICMU2010@Seattle<br />
  26. 26. We developed own simulator in Java<br />Simulation Configuration<br />Simulation parameters<br />User mobility model<br />2010/4/26<br />23<br />ICMU2010@Seattle<br />
  27. 27. Total Satisfaction<br />Overall importance scores associate with data that were delivered within deadline.<br />A larger value means a better performance<br />We confirmed total satisfaction by adjusting required delivery ratio δ<br />Comparison queuing methods<br />Ⅰ. FIFO : First in first out<br />Ⅱ. Satisfaction: Sorted by satisfaction point<br />Ⅲ. Deadline : Sorted by deadline<br />Metric<br />2010/4/26<br />24<br />ICMU2010@Seattle<br />
  28. 28. Total Satisfactionvs. Required delivery ratio(δ)<br />Result:<br /><ul><li>A higherδ leads a higher total satisfaction
  29. 29. Proposed method is delivery C-P sensitive
  30. 30. Achieved significant improvement</li></ul>δ<br />2010/4/26<br />25<br />ICMU2010@Seattle<br />
  31. 31. Conclusion<br />Proposed a method to maximize overall user satisfaction for data delivery in congested DTN environments<br />Confirmed good effect of the proposed method by comparing with 3 conventional methods<br />Future Work<br />Compare with Epidemic method, etc.<br />Find some mechanisms to reduce network load when congestion occurs<br />Conclusion<br />2010/4/26<br />26<br />ICMU2010@Seattle<br />
  32. 32. Thank You!<br />Any Questions?<br />2010/4/26<br />27<br />ICMU2010@Seattle<br />
  33. 33. 2010/4/26<br />ICMU2010@Seattle<br />28<br />
  34. 34. Delivery time estimationfor each data<br />ε<br />γ<br />β<br />Reply delivery time<br />Target Spot<br />Receiving Spot<br />Source Spot<br />RS staying time<br />User traveling time<br />How to estimate deadline<br />User traveling time (γ) is from posting request at Source Spot until he/she arrives at Receiving Spot<br />RS staying time (ε) is staying time when user is at Receiving Spot <br />The total time of γ and ε is Deadline<br />How to estimate data delivery time<br />Request delivery time α is the delivery time from a request was posted at Source Spot until it was carried to Target Spot<br />Reply delivery timeβ is the traveling time until user arrives at RS<br />α<br />Request delivery time<br />Deadline = γ+ε<br />Deadline ≧ α+β<br />2010/4/26<br />29<br />ICMU2010@Seattle<br />
  35. 35. 期待到達率(δ) vs 平均期待コストパフォーマンス(ECP)<br />Result: <br /><ul><li>期待到達率を向上させるために、データの複製数を増やさなければならない
  36. 36. データ複製数の増加は、ECPの低下を引き起こす
  37. 37. 同値の期待到達率において、提案手法のECPが最も高い</li></ul>2010/4/26<br />30<br />ICMU2010@Seattle<br />
  38. 38. 配送データの満足度総和を最大化するために,以下の課題を解決する必要がある<br />配送時間の見積もり<br />受信期限に間に合わないデータによる通信容量の浪費を軽減<br />適切なデータ複製数の算出<br />到達率向上させるためにデータを複製<br />過度複製によるオーバヘッドの回避<br />配送コストパフォーマンスの算出<br />配送効率の良いデータを優先的に配送<br />情報Boxにこれらを解決する機能を実装<br />Target Problems<br />2010/4/26<br />31<br />ICMU2010@Seattle<br />
  39. 39. Restrain performance reduction<br />Use high c-p data delivery to save bandwidth resource<br />Why calculate data cost-performance?<br />Importance score<br />60<br />30<br />30<br />30<br />Send<br />Data size<br />60<br />30<br />30<br />30<br />Expired<br />60<br />30<br />30<br />30<br />Discard<br />30<br />30<br />30<br />Higher overall satisfaction <br />is achieved<br />2010/4/26<br />32<br />ICMU2010@Seattle<br />

2010/04/05 ICMU@Seattle

Views

Total views

3,097

On Slideshare

0

From embeds

0

Number of embeds

2,056

Actions

Downloads

0

Shares

0

Comments

0

Likes

0

×