Traitement en cours

Veuillez attendre...

Paramétrages

Paramétrages

Aller à Demande

1. WO2020111989 - ADAPTATION AUTOMATIQUE ET DYNAMIQUE DE REGROUPEMENT DANS UN SYSTÈME DE TRAITEMENT DE DONNÉES

Note: Texte fondé sur des processus automatiques de reconnaissance optique de caractères. Seule la version PDF a une valeur juridique

[ EN ]

IN THE CLAIMS:

1. A first leader node (320, 600) of a first group of nodes (310) in

communication with a plurality of leader nodes (350, 380) in a distributed data processing system (300), comprising:

processing circuitry (610), configured to:

determine (815) a cost map including computational and/or communication costs within said first group of nodes (310);

calculate (820) costs between said first leader node (320, 600) and first member nodes (330) within said first group of nodes (310) based on said cost map;

determine (830) a candidate member node (330 A) from said first member nodes (330) if said costs between said first leader node (320) and said candidate member node (330 A) exceeds a threshold;

query (840) said plurality of leader nodes (350, 380) to accept a transfer of said candidate member node (330A);

receive (850) a first transfer response from a second leader node (350) of said plurality of leader nodes (350, 380) to accept said transfer of said candidate member node (330A) to a second group of nodes (340) directed by said second leader node (350);

initiate (860) a transfer of said candidate member node (330 A) to said second group of nodes (340) if said first leader node (320, 600) accepts said transfer thereto; and

receive (865) an acknowledgement of said transfer of said candidate member node (330A) to said second group of nodes (340).

2. The first leader node (320, 600) as recited in Claim 1 wherein said processing circuitry (610) is configured to update (810) membership information and status for said first member nodes (330) within said first group of nodes (310).

3. The first leader node (320, 600) as recited in Claim 1 wherein said processing circuitry (610) is configured to compare (825) said costs between said first leader node (320, 600) and said first member nodes (330).

4. The first leader node (320, 600) as recited in Claim 1 wherein said query comprises an identifier for said candidate member node (330A) and said costs.

5. The first leader node (320, 600) as recited in Claim 1 wherein said second leader node (350) takes into account costs and ease of adaptation to transfer said candidate member node (330A) to said second group of nodes (340).

6. The first leader node (320, 600) as recited in Claim 1 wherein said processing circuitry (610) is configured to set (835) a timer and check for said first transfer response when said timer expires.

7. The first leader node (320, 600) as recited in Claim 1 wherein said processing circuitry (610) is configured to initiate (860) said transfer of said candidate member node (330A) to said second group of nodes (340) in accordance with a join message to said second leader node (350) from said candidate member node (330A).

8. The first leader node (320, 600) as recited in Claim 1 wherein said processing circuitry (610) is configured to receive (875) a second transfer response from a third leader node (380) of said plurality of leader nodes (350, 380) to accept said transfer of said candidate member node (330A) to a third group of nodes (370) directed by said third leader node (380).

9. The first leader node (320, 600) as recited in Claim 8 wherein said processing circuitry (610) is configured to:

initiate (885) a transfer of said candidate member node (330A) to said third group of nodes (370) if said first leader node (320, 600) accepts said transfer thereto; and

receive (890) an acknowledgement of said transfer of said candidate member node (330A) to said third group of nodes (370).

10. The first leader node (320, 600) as recited in Claim 8 wherein said processing circuitry (610) is configured to set (835) a timer and check for said first transfer response and said second transfer response when said timer expires.

11. The first leader node (320, 600) as recited in Claim 1 wherein said cost map takes into account a network topology of said distributed data processing system (300).

12. The first leader node (320, 600) as recited in Claim 1 wherein said costs comprise at least one of communication costs between said first leader node (320, 600) and said first member nodes (330), physical proximity between said first leader node (320, 600) and said first member nodes (330), and hardware similarity between said first leader node (320, 600) and said first member nodes (330).

13. The first leader node (320, 600) as recited in Claim 1 wherein said first leader node (320, 600) is also a member of another group of nodes (395) in a higher layer of said distributed data processing system (300).

14. The first leader node (320, 600) as recited in Claim 1 wherein said processing circuitry (610) is also configured to determine (830) said candidate member node (330A) based on a size of said first group of nodes (310).

15. A method (800) of operating a first leader node (320, 600) of a first group of nodes (310) in communication with a plurality of leader nodes (350, 380) in a distributed data processing system (300), comprising:

determining (815) a cost map including computational and/or communication costs within said first group of nodes (310);

calculating (820) costs between said first leader node (320, 600) and first member nodes (330) within said first group of nodes (310) based on said cost map;

determining (830) a candidate member node (330 A) from said first member nodes (330) if said costs between said first leader node (320) and said candidate member node (330A) exceeds a threshold;

querying (840) said plurality of leader nodes (350, 380) to accept a transfer of said candidate member node (330 A);

receiving (850) a first transfer response from a second leader node (350) of said plurality of leader nodes (350, 380) to accept said transfer of said candidate member node (330A) to a second group of nodes (340) directed by said second leader node (350); initiating (860) a transfer of said candidate member node (330 A) to said second group of nodes (340) if said first leader node (320, 600) accepts said transfer thereto; and receiving (865) an acknowledgement of said transfer of said candidate member node (330A) to said second group of nodes (340).

16. The method (800) as recited in Claim 15 further comprising updating (810) membership information and status for said first member nodes (330) within said first group of nodes (310).

17. The method (800) as recited in Claim 15 further comprising comparing (825) said costs between said first leader node (320, 600) and said first member nodes (330).

18. The method (800) as recited in Claim 15 wherein said querying (840) comprises information including an identifier for said candidate member node (330A) and said costs.

19. The method (800) as recited in Claim 15 wherein said second leader node (350) takes into account costs and ease of adaptation to transfer said candidate member node (330A) to said second group of nodes (340).

20. The method (800) as recited in Claim 15 further comprising setting (835) a timer and checking for said first transfer response when said timer expires.

21. The method (800) as recited in Claim 15 further comprising initiating (860) said transfer of said candidate member node (330A) to said second group of nodes (340) in accordance with a join message to said second leader node (350) from said candidate member node (330A).

22. The method (800) as recited in Claim 15 further comprising receiving (875) a second transfer response from a third leader node (380) of said plurality of leader nodes (350, 380) to accept said transfer of said candidate member node (330A) to a third group of nodes (370) directed by said third leader node (380).

23. The method (800) as recited in Claim 22, further comprising:

initiating (885) a transfer of said candidate member node (330A) to said third group of nodes (370) if said first leader node (320, 600) accepts said transfer thereto; and receiving (890) an acknowledgement of said transfer of said candidate member node (330A) to said third group of nodes (370).

24. The method (800) as recited in Claim 22 further comprising setting (835) a timer and checking for said first transfer response and said second transfer response when said timer expires.

25. The method (800) as recited in Claim 15 wherein said cost map takes into account a network topology of said distributed data processing system (300).

26. The method (800) as recited in Claim 15 wherein said costs comprise at least one of communication costs between said first leader node (320, 600) and said first member nodes (330), physical proximity between said first leader node (320, 600) and said first member nodes (330), and hardware similarity between said first leader node (320, 600) and said first member nodes (330).

27. The method (800) as recited in Claim 15 wherein said first leader node (320, 600) is also a member of another group of nodes (395) in a higher layer of said distributed data processing system (300).

28. The method (800) as recited in Claim 15 wherein said determining (830) said candidate member node (330A) is based on a size of said first group of nodes (310).

29. A first leader node (420, 600) of a first group of nodes (410) in communication with a second leader node (450) of a second group of nodes (440) in a distributed data processing system (400), comprising:

processing circuitry (610), configured to:

receive (915) a query from said second leader node (450) to accept a transfer of a candidate member node (460A) of second member nodes (460) of said second group of nodes (440) to said first group of nodes (410);

provide (950) a transfer response to said second leader node (450) if said first leader node (420, 600) accepts said transfer of said candidate member node (460A) to said first group of nodes (410);

receive (955) an instruction to transfer said candidate member node (460 A) to said first group of nodes (410); and

provide (960) an acknowledgement of said transfer of said candidate member node (460a) to said first group of nodes (410).

30. The first leader node (420, 600) as recited in Claim 29 wherein said processing circuitry (610) is configured to:

calculate (925) costs of said candidate member node (460A) within said first group of nodes (410) based on a cost map including computational and/or communication costs within said first group of nodes (410); and

provide (940) said transfer response to accept said transfer if said costs between said first leader node (420, 600) and said candidate member node (460A) within said first group of nodes (410) are less than a threshold.

31. The first leader node (420, 600) as recited in Claim 30 wherein said processing circuitry (610) is configured to compare (930) said costs between said first leader node (420, 600) and said candidate member node (460A) within said first group of nodes (410).

32. The first leader node (420, 600) as recited in Claim 30 wherein said cost map takes into account a network topology of said distributed data processing system (400).

33. The first leader node (420, 600) as recited in Claim 30 wherein said costs comprise at least one of communication costs between said first leader node (420, 600) and said candidate member node (460A), physical proximity between said first leader node (420, 600) and said candidate member node (460A), and hardware similarity between said first leader node (420, 600) and said candidate member node (460A).

34. The first leader node (420, 600) as recited in Claim 29 wherein said processing circuitry (610) is configured to update (910) membership information and status for first member nodes (430) within said first group of nodes (410).

35. The first leader node (420, 600) as recited in Claim 29 wherein said query comprises an identifier for said candidate member node (460A) and costs to remain in said second group of nodes (440).

36. The first leader node (420, 600) as recited in Claim 29 wherein said first leader node (420, 600) takes into account costs and ease of adaptation to transfer said candidate member node (460A) to said first group of nodes (410).

37. The first leader node (420, 600) as recited in Claim 29 wherein said instruction to transfer said candidate member node (460A) to said first group of nodes (410) is in accordance with a join message from said candidate member node (460A).

38. The first leader node (420, 600) as recited in Claim 29 wherein said first leader node (420, 600) is also a member of another group of nodes (495) in a higher layer of said distributed data processing system (400).

39. The first leader node (420, 600) as recited in Claim 29 wherein said processing circuitry (610) is configured to provide (950) said transfer response to accept said transfer based on a size of said first group of nodes (410).

40. A method (900) of operating a first leader node (420, 600) of a first group of nodes (410) in communication with a second leader node (450) of a second group of nodes (440) in a distributed data processing system (400), comprising:

receiving (915) a query from said second leader node (450) to accept a transfer of a candidate member node (460A) of second member nodes (460) of said second group of nodes (440) to said first group of nodes (410);

providing (950) a transfer response to said second leader node (450) if said first leader node (420, 600) accepts said transfer of said candidate member node (460A) to said first group of nodes (410);

receiving (955) an instruction to transfer said candidate member node (460 A) to said first group of nodes (410); and

providing (960) an acknowledgement of said transfer of said candidate member node (460a) to said first group of nodes (410).

41. The method (900) as recited in Claim 40, further comprising:

calculating (925) costs of said candidate member node (460A) within said first group of nodes (410) based on a cost map including computational and/or communication costs within said first group of nodes (410); and

providing (940) said transfer response to accept said transfer if said costs between said first leader node (420, 600) and said candidate member node (460A) within said first group of nodes (410) are less than a threshold.

42. The method (900) as recited in Claim 41 further comprising comparing (930) said costs between said first leader node (420, 600) and said candidate member node (460A) within said first group of nodes (410).

43. The method (900) as recited in Claim 41 wherein said cost map takes into account a network topology of said distributed data processing system (400).

44. The method (900) as recited in Claim 41 wherein said costs comprise at least one of communication costs between said first leader node (420, 600) and said candidate member node (460A), physical proximity between said first leader node (420, 600) and said candidate member node (460A), and hardware similarity between said first leader node (420, 600) and said candidate member node (460A).

45. The method (900) as recited in Claim 40 further comprising updating (910) membership information and status for first member nodes (430) within said first group of nodes (410).

46. The method (900) as recited in Claim 40 wherein said query comprises an identifier for said candidate member node (460A) and costs to remain in said second group of nodes (440).

47. The method (900) as recited in Claim 40 wherein said first leader node (420, 600) takes into account costs and ease of adaptation to transfer said candidate member node (460 A) to said first group of nodes (410).

48. The method (900) as recited in Claim 40 wherein said instruction to transfer said candidate member node (460 A) to said first group of nodes (410) is in accordance with a join message from said candidate member node (460A).

49. The method (900) as recited in Claim 40 wherein said first leader node (420, 600) is also a member of another group of nodes (495) in a higher layer of said distributed data processing system (400).

50. The method (900) as recited in Claim 40 wherein said processing circuitry

(610) is configured to provide (950) said transfer response to accept said transfer based on a size of said first group of nodes (410).