Opportunistic composition of sequentially-connected services in mobile computing environments

878 views
789 views

Published on

Presentation at the International Conference for Web Services 2011. Presented with Best Student Paper Award.

Imagine using mobile device capabilities and participatory sensing to create new value-added services from existing device services

Published in: Technology, Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
878
On SlideShare
0
From Embeds
0
Number of Embeds
5
Actions
Shares
0
Downloads
3
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Opportunistic composition of sequentially-connected services in mobile computing environments

  1. 1. Opportunistic composition of sequentially-connected services in mobile computing environments Christin Groba and Siobhán Clarke – Trinity College Dublin IEEE International Conference of Web Services Washington D.C., 5th July 2011Lero© 2011
  2. 2. Mobile environments are the new frontier for service providersLero© 2011 2
  3. 3. Getting to the airportLero© 2011 3
  4. 4. Fastest routeLero© 2011 4
  5. 5. DisruptionLero© 2011 5
  6. 6. What is the fastest alternative ?Lero© 2011 6
  7. 7. What is the fastest alternative ? Browsing the Web Walk 5 min Walk 3 min Bus in 6 min Reality Walk 5 minLero© 2011 7
  8. 8. How to obtain real-time data? GPS Accel Navi Camera Utilise mobile device capabilities & participatory sensingLero© 2011 8
  9. 9. Real-time data as decision support Real-time data Walk 5 min Real-time data Walk 5 minLero© 2011 9
  10. 10. Mobility and resource scarcity make service composition challengingLero© 2011 10
  11. 11. Challenge: Dynamic setting No central control High failure probabilityLero© 2011 11
  12. 12. Challenge: Energy scarcity Wireless communication drains batteryLero© 2011 12
  13. 13. Emergence of dynamic ad hoc environments has led to decentralised composition algorithmsLero© 2011 13
  14. 14. Decentralised composition BindingLero© 2011 14
  15. 15. Decentralised composition BindingLero© 2011 15
  16. 16. Decentralised composition BindingLero© 2011 16
  17. 17. Decentralised composition BindingLero© 2011 17
  18. 18. Decentralised composition BindingLero© 2011 18
  19. 19. Decentralised composition BindingLero© 2011 19
  20. 20. Decentralised composition Binding Once all service are bound, execution can startLero© 2011 20
  21. 21. Decentralised composition Execution Binding Once all service are bound, execution can startLero© 2011 21
  22. 22. Decentralised composition Execution Binding Once all service are bound, execution can startLero© 2011 22
  23. 23. Decentralised composition Execution Binding Once all service are bound, execution can startLero© 2011 23
  24. 24. Decentralised composition Execution Binding Once all service are bound, execution can startLero© 2011 24
  25. 25. Decentralised composition Execution Binding Once all service are bound, execution can startLero© 2011 25
  26. 26. Decentralised composition Execution Binding Once all service are bound, execution can startLero© 2011 26
  27. 27. Many decentralised composition approaches bind providers too earlyLero© 2011 27
  28. 28. Bound but not usedLero© 2011 28
  29. 29. Disconnect Recovery Communication EnergyLero© 2011 29
  30. 30. How to reduce failure probability?Lero© 2011 30
  31. 31. Reduce the delay between service binding and service executionLero© 2011 31
  32. 32. Opportunistic execution model BindingLero© 2011 32
  33. 33. Opportunistic execution model Execution Binding Execute service while searching for the next providerLero© 2011 33
  34. 34. Opportunistic execution model Execution Binding Execute service while searching for the next providerLero© 2011 34
  35. 35. Opportunistic execution model Execution Binding Execute service while searching for the next providerLero© 2011 35
  36. 36. Opportunistic execution model Execution Binding No unnecessary binding of conditional pathsLero© 2011 36
  37. 37. Opportunistic execution model Execution BindingLero© 2011 37
  38. 38. Opportunistic execution model Execution Binding Bind available providerLero© 2011 38
  39. 39. Opportunistic execution model Execution Binding Provider likely to be still available if execution follows immediately after bindingLero© 2011 39
  40. 40. Opportunistic execution model Execution BindingLero© 2011 40
  41. 41. System model Abstract composite request < composite > := seq(< composite >;< composite >) |< type > | < type >:< provider > < type > := S1 |…| Sn < provider > := P1 |…| Pm Directory-less, on demand discovery Proximity-based selectionLero© 2011 41
  42. 42. Evaluation Simulation Jist/SWANS Ulm Random Waypoint Controlled Composition length Node speed Metric Success ratio Response time Communication effortLero© 2011 42
  43. 43. Do composites execute to completion? Opportunistic model 28% more successful Node speed: 1-13 m/s 80 60 Success ratio(%) 40 baseline opportunistic 20 0 4 5 6 7 Composition lengthLero© 2011 43
  44. 44. Do composites respond to initiator quicker? Opportunistic model 33% faster Node speed: 1-13 m/s 8 Response time (s) 6 baseline 4 opportunistic 2 0 4 5 6 7 Composition lengthLero© 2011 44
  45. 45. Do composites communicate unnecessarily? Opportunistic model 14% less chatty Node speed: 1-13 m/s 120 Number of sent messages 100 80 60 baseline opportunistic 40 20 0 4 5 6 7 Composition lengthLero© 2011 45
  46. 46. Current & Future work Service flows Service requests contain splitting and merging paths Failure recovery Composition success cannot be guaranteed and recovery strategies are essential Privacy Participants aware of control and data flow which may include sensitive information Device heterogeneity Resource-poor devices require light-weight integration conceptLero© 2011 46
  47. 47. Recap New frontier Share mobile device capabilities and create new value-added service composites Challenge How to reduce the failure probability of service composites? Approach Reduce delay between binding and execution Results More successful, faster, and less chatty than “first bind all then execute” approach Current work Service flows and failure recovery Future work Privacy and device heterogeneityLero© 2011 47
  48. 48. Thank you!Lero© 2011

×