1 / 22

BreadCrumbs: Forecasting Mobile Connectivity

Presented by Hao He. BreadCrumbs: Forecasting Mobile Connectivity. Anthony J. Nicholson and Brian D. Noble. Slides adapted from Dhruv Kshatriya. Observations. Access points come and go as users move Not all network connections created equal Limited time to exploit a given connection.

sugar
Download Presentation

BreadCrumbs: Forecasting Mobile Connectivity

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Presented by Hao He BreadCrumbs: Forecasting Mobile Connectivity Anthony J. Nicholson and Brian D. Noble Slides adapted from DhruvKshatriya

  2. Observations Access points come and go as users move Not all network connections created equal Limited time to exploit a given connection

  3. The big idea(s) in this paper‏ Introduce the concept of connectivity forecasts Show how such forecasts can be accurate for everyday situations w/o GPS or centralization Illustrate through example applications

  4. Road Map • Background knowledge • Connectivity forecasting • Evaluation • Conclusion

  5. Background knowledge • Determining AP quality • Wifi-Reports: • Improving Wireless Network Selection with Collaboration • Estimating Client Location

  6. Improved Access Point Selection Conventionally AP’s with the highest signal strength are chosen. Probe application-level quality of access points • Bandwidth, latency, open ports • AP quality database guides future selection Real-world evaluation • Significant improvement over link-layer metrics

  7. Determining location • Best: GPS on device • Unreasonable assumption? • PlaceLab • Triangulate 802.11 beacons • Wardriving databases • Other options • Accelerometer, GSM beacons

  8. Connectivity Forecasting • Maintain a personalized mobility modelon the user's device to predict future associations • Combine prediction with AP quality database to produce connectivity forecasts • Applications use these forecasts to take domain-specific actions

  9. Mobility model Humans are creatures of habit • Common movement patterns • Second-order Markov chain • Reasonable space and time overhead (mobile device)‏ • Literature shows as effective as fancier methods • State: current GPS coord + last GPS coord • Coords rounded to one-thousandth of degree(110m x 80m box)

  10. Mobility model example

  11. Connectivity forecasts Applications and kernel query BreadCrumbs Expected bandwidth (or latency, or...) in the future Recursively walk tree based on transition frequency

  12. 0.17 0.22 0.61 Forecast example: downstream BW What will the available downstream bandwidthbe in 10 seconds (next step)? 0.61*72.13 + 0.17*0.00 + 0.22*141.84 = 75.20 KB/s current 72.13 0.00 141.84

  13. Evaluation methodology • Tracked weekday movements for two weeks • Linux 2.6 on iPAQ + WiFi • Mixture of walking, driving, and bus • Primarily travel to/from office, but some noise • Driving around for errands • Walk to farmers' market, et cetera • Week one as training set, week two for eval

  14. AP statistics

  15. Forecast accuracy

  16. Application: opportunistic writeback

  17. Application: Radio Deactivation Goal • Conserving energy Implementation • Query BreadCrumbs to get a connectivity forecast • If radio on & no connectivity in next 30 secs Turn radio off • Else If radio off & BreadCrumbs predicts connectivity in next 30 secs

  18. Application: Radio Deactivation

  19. Application: Phone network vs. WiFi

  20. Summary Humans (and their devices) are creatures of habit Mobility model + AP quality DB = connectivity forecasts Minimal application modifications yield benefits to user

  21. Future work Evaluation: not representative Energy efficient Modification to software Limited to certain applications: ex. download

  22. Thank you!

More Related