PES: Proactive Event Scheduling for Responsive and Energy-Efficient Mobile Web Computing

ISCA 2019 Talk. Presented by Yu Feng

Transcript

1. PES: Proactive Event Scheduling
   for Responsive and Energy-Efficient
   
   Mobile Web Computing
   Yu Feng with Yuhao Zhu
   
   Department of Computer Science
   
   University of Rochester
   
   ISCA 2019
   

   View Slide

2. Ubiquity of Mobile devices
   2
   

   View Slide

3. Ubiquity of Mobile devices
   2
   

   View Slide

4. Ubiquity of Mobile devices
   3
   

   View Slide

5. Ubiquity of Mobile devices
   3
   ▸2x times mobile users than
   desktop users
   

   View Slide

6. Ubiquity of Mobile devices
   3
   ▸2x times mobile users than
   desktop users
   ▸76% of population are using
   mobile devices daily
   

   View Slide

7. Mobile Applications are Event-Driven
   4
   

   View Slide

8. Mobile Applications are Event-Driven
   4
   

   View Slide

9. Mobile Applications are Event-Driven
   4
   Typing
   Interactions
   ISCA 2019
   

   View Slide

10. Mobile Applications are Event-Driven
    4
    Scrolling
    Typing
    Interactions
    ISCA 2019
    

    View Slide

11. Mobile Applications are Event-Driven
    4
    Scrolling
    Tapping
    Typing
    Interactions
    ISCA 2019
    

    View Slide

12. Mobile Applications are Event-Driven
    4
    Scrolling
    Tapping
    scrollstart
    scrollupdate
    scrollend
    tapdown
    Tapup
    Keypress
    Typing
    Interactions Events
    ISCA 2019
    

    View Slide

13. View Slide

14. View Slide

15. Event-Driven Execution Model
    6
    Event
    

    View Slide

16. Event-Driven Execution Model
    6
    touchmove
    onclick
    timer
    …
    Time
    Event Queue
    Push
    Event
    

    View Slide

17. Event-Driven Execution Model
    7
    touchmove
    onclick
    timer
    …
    Time
    Event Queue
    Fetch
    Runtime
    Scheduler
    Core
    

    View Slide

18. Event-Driven Execution Model
    7
    touchmove
    onclick
    timer
    …
    Time
    Event Queue
    Fetch
    Runtime
    Scheduler
    Core
    Scheduling knobs:
    
    DVFS settings
    

    View Slide

19. Inefficiency of Current Schedulers
    8
    

    View Slide

20. Inefficiency of Current Schedulers
    8
    Time
    

    View Slide

21. Inefficiency of Current Schedulers
    8
    Time
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    

    View Slide

22. Inefficiency of Current Schedulers
    8
    Time
    OS Governor
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    

    View Slide

23. Inefficiency of Current Schedulers
    8
    Time
    OS Governor
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1
    

    View Slide

24. Inefficiency of Current Schedulers
    8
    Time
    OS Governor
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 slack
    

    View Slide

25. Inefficiency of Current Schedulers
    8
    Time
    OS Governor
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    slack
    

    View Slide

26. Inefficiency of Current Schedulers
    9
    Time
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    OS Governor
    

    View Slide

27. Inefficiency of Current Schedulers
    9
    Time
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    QoS-Aware
    OS Governor
    

    View Slide

28. Inefficiency of Current Schedulers
    9
    Time
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    E1
    QoS-Aware
    OS Governor
    

    View Slide

29. Inefficiency of Current Schedulers
    9
    Time
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    E1 E2
    QoS-Aware
    OS Governor
    

    View Slide

30. Inefficiency of Current Schedulers
    9
    Time
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    E1 E2
    QoS-Aware
    OS Governor
    ?
    

    View Slide

31. Inefficiency of Current Schedulers
    10
    Time
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    E1 E2
    QoS-Aware
    OS Governor
    

    View Slide

32. Inefficiency of Current Schedulers
    10
    Time
    Oracle
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    E1 E2
    QoS-Aware
    OS Governor
    

    View Slide

33. Inefficiency of Current Schedulers
    10
    Time
    Oracle
    input 1 input 2
    QoS
    
    Deadline 1
    QoS
    
    Deadline 2
    E1 E2
    E1 E2
    E1 E2
    QoS-Aware
    OS Governor
    

    View Slide

34. Inefficiency of Current Schedulers
    11
    Time
    E1 E2 E3 E4 E5
    Schedule across events
    

    View Slide

35. Inefficiency of Current Schedulers
    12
    Time
    E1 E2 E3 E4 E5
    Current
    events
    

    View Slide

36. Inefficiency of Current Schedulers
    12
    Time
    E1 E2 E3 E4 E5
    Current
    events
    Future events
    

    View Slide

37. Proactive Event Scheduler
    13
    

    View Slide

38. Proactive Event Scheduler
    Proactive Event Scheduler
    13
    

    View Slide

39. Proactive Event Scheduler
    Proactive Event Scheduler
    13
    Prediction
    Web Program
    
    Analysis
    Machine
    
    Learning
    

    View Slide

40. Proactive Event Scheduler
    Proactive Event Scheduler
    13
    Prediction
    Web Program
    
    Analysis
    Machine
    
    Learning
    Scheduling
    Constraint Optimization
    

    View Slide

41. Proactive Event Scheduler
    Proactive Event Scheduler
    13
    Result reduce energy consumption while improving QoS
    Prediction
    Web Program
    
    Analysis
    Machine
    
    Learning
    Scheduling
    Constraint Optimization
    

    View Slide

42. Proactive Event Scheduler
    Proactive Event Scheduler
    13
    Result reduce energy consumption while improving QoS
    Prediction
    Web Program
    
    Analysis
    Machine
    
    Learning
    Scheduling
    Constraint Optimization
    

    View Slide

43. Event Sequence Learning
    14
    Prediction
    Model
    

    View Slide

44. Event Sequence Learning
    14
    Time
    E1 E2 E3
    Event Sequence
    Prediction
    Model
    

    View Slide

45. Event Sequence Learning
    14
    Time
    E1 E2 E3 ?
    Event Sequence
    Prediction
    Model
    

    View Slide

46. Event Sequence Learning
    14
    Time
    E1 E2 E3 ?
    Event Sequence
    Prediction
    Model
    

    View Slide

47. Event Sequence Learning
    14
    Time
    E1 E2 E3 ?
    Event Sequence
    E4
    Prediction
    Model
    

    View Slide

48. Event Sequence Learning
    14
    Time
    E1 E2 E3 ?
    Event Sequence
    E4
    Prediction
    Model
    

    View Slide

49. Recurrent Prediction
    15
    Time
    E1 E2 E3 E4
    Event Sequence
    Prediction
    Model
    

    View Slide

50. Recurrent Prediction
    15
    Time
    E1 E2 E3 E4
    Event Sequence
    Prediction
    Model
    

    View Slide

51. Recurrent Prediction
    15
    Time
    E1 E2 E3 E4
    Event Sequence
    E5
    Prediction
    Model
    

    View Slide

52. Recurrent Prediction
    15
    Time
    E1 E2 E3 E4
    Event Sequence
    E5
    Prediction
    Model
    

    View Slide

53. Recurrent Prediction
    15
    Time
    E1 E2 E3 E4
    Event Sequence
    E5
    …
    Prediction
    Model
    

    View Slide

54. Prediction Model
    16
    Prediction
    Model
    

    View Slide

55. Prediction Model
    17
    Prediction
    Model
    

    View Slide

56. Prediction Model
    17
    Prediction
    Model
    The distance of click
    The number of scrolls
    The number of navigations
    Features
    
    encoding
    
    past
    interactions
    …
    

    View Slide

57. Prediction Model
    18
    Features
    
    encoding
    
    past
    
    interactions
    ln(
    p
    1 − p
    ) =
    x
    β
    Prediction
    Model
    

    View Slide

58. Prediction Model
    19
    Features
    
    encoding
    
    past
    
    interactions
    Click
    ScrollUp
    ScrollDown
    ZoomIn
    ZoomOut
    …
    Prediction
    Model
    0.10
    0.12
    0.58
    0.07
    0.02
    

    View Slide

59. Prediction Model
    19
    Features
    
    encoding
    
    past
    
    interactions
    Click
    ScrollUp
    ScrollDown
    ZoomIn
    ZoomOut
    …
    Prediction
    Model
    ✔
    

    View Slide

60. Prediction Model
    19
    Features
    
    encoding
    
    past
    
    interactions
    Click
    ScrollUp
    ScrollDown
    ZoomIn
    ZoomOut
    …
    Prediction
    Model
    All
    
    Event Types
    

    View Slide

61. Prediction Model
    19
    Features
    
    encoding
    
    past
    
    interactions
    Click
    ScrollUp
    ScrollDown
    ZoomIn
    ZoomOut
    …
    Prediction
    Model
    Filter
    
    Events
    ?
    

    View Slide

62. 20
    Program Analysis
    

    View Slide

63. 20
    Program Analysis
    

    View Slide

64. 20
    Program Analysis
    

    View Slide

65. 20
    Program Analysis
    Viewport
    

    View Slide

66. 20
    
    
    
    
    
    …
    
    
    
    
    …
    Program Analysis
    DOM Tree
    Viewport
    

    View Slide

67. 20
    
    
    
    
    
    …
    
    
    
    
    …
    Viewport
    Program Analysis
    DOM Tree
    Viewport
    

    View Slide

68. 20
    
    
    
    
    
    …
    
    
    
    
    …
    Viewport
    Program Analysis
    DOM Tree
    Viewport
    

    View Slide

69. 20
    
    
    
    
    
    …
    
    
    
    
    …
    Viewport
    Program Analysis
    DOM Tree
    Viewport
    

    View Slide

70. 20
    
    
    
    
    
    …
    
    
    
    
    …
    Viewport
    Program Analysis
    DOM Tree
    Viewport
    

    View Slide

71. Include Program Analysis Features
    21
    Features
    
    encoding
    
    past
    
    interactions
    Click
    ScrollUp
    ScrollDown
    ZoomIn
    ZoomOut
    …
    Prediction
    Model
    

    View Slide

72. Overview of Predictor
    22
    Past Event
    
    Statistics
    Current Application
    
    State
    Click
    ScrollUp
    ScrollDown
    ZoomIn
    ZoomOut
    …
    +
    Prediction
    Model
    

    View Slide

73. Overview of Predictor
    22
    Past Event
    
    Statistics
    Current Application
    
    State
    Click
    ScrollUp
    ScrollDown
    ZoomIn
    ZoomOut
    …
    +
    Prediction
    Model
    Recurrent
    

    View Slide

74. Overview of Predictor
    22
    

    View Slide

75. Overview of Predictor
    22
    High Accuracy, low overhead!
    

    View Slide

76. Proactive Event Scheduler
    Proactive Event Scheduler
    23
    Result reduce energy consumption meanwhile improve QoS
    Prediction
    Web Program
    
    Analysis
    Machine
    
    Learning
    Scheduling
    Constraint Optimization
    

    View Slide

77. Proactive Event Scheduler
    Proactive Event Scheduler
    23
    Result reduce energy consumption meanwhile improve QoS
    Prediction
    Web Program
    
    Analysis
    Machine
    
    Learning
    Scheduling
    Constraint Optimization
    

    View Slide

78. Scheduling events
    24
    E1 E2 E3
    Time
    From
    
    predictor
    

    View Slide

79. Scheduling events
    24
    E1 E2 E3
    Time
    From
    
    predictor
    Goal: minimize total energy while meeting deadlines
    

    View Slide

80. Problem Formulation
    25
    E1 E2 E3
    Time
    ▸ Scheduling Problem → Constrained Optimization.
    

    View Slide

81. Objective:
    Problem Formulation
    25
    E1 E2 E3
    Time
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    Min. Energy (i)
    

    View Slide

82. Objective:
    Problem Formulation
    25
    E1 E2 E3
    Time
    △Texe 1 △Texe 2 △Texe 3
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x
    Min.
    

    View Slide

83. Objective:
    Problem Formulation
    25
    E1 E2 E3
    Time
    △Texe 1 △Texe 2 △Texe 3
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    

    View Slide

84. Objective:
    Constraints:
    Problem Formulation
    25
    E1 E2 E3
    Time
    △Texe 1 △Texe 2 △Texe 3
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    

    View Slide

85. Objective:
    Constraints:
    Problem Formulation
    25
    E1 E2 E3
    Time
    △Texe 1 △Texe 2 △Texe 3
    ▸ Scheduling Problem → Constrained Optimization.
    Order: ≤
    Tend (i) Tstart (i+1)
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    

    View Slide

86. Objective:
    Constraints:
    Problem Formulation
    25
    E1 E2 E3
    Time
    Tstart 1 Tstart 2 Tstart 3
    TQoS 1 TQoS 2 TQoS 3
    △Texe 1 △Texe 2 △Texe 3
    ▸ Scheduling Problem → Constrained Optimization.
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    

    View Slide

87. Objective:
    Constraints:
    Problem Formulation
    25
    E1 E2 E3
    Time
    Tstart 1 Tstart 2 Tstart 3
    TQoS 1 TQoS 2 TQoS 3
    △Texe 1 △Texe 2 △Texe 3
    ▸ Scheduling Problem → Constrained Optimization.
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    Scheduling knobs:
    
    DVFS settings
    
    for each event
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    

    View Slide

88. Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

89. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    △Texe (i) =
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

90. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    △Texe (i) = Tmemory
    +
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

91. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    Tcpu
    △Texe (i) = Tmemory
    +
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

92. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    △Texe (i) = Tmemory
    + Ncycles / f (i)
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

93. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    △Texe (i) = Tmemory
    + Constants
    Ncycles / f (i)
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

94. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Power (i)
    Min.
    △Texe (i) = Tmemory
    + Constants
    Offline profile
    Ncycles / f (i)
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

95. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    26
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x
    Min.
    △Texe (i) = Tmemory
    + Constants
    Offline profile
    Ncycles / f (i)
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    Pmap (i)
    

    View Slide

96. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    27
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Pmap (i)
    Min.
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    

    View Slide

97. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    27
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Pmap (i)
    Min.
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    { Ncycles / f (j) }
    △Texe (i) =Tmemory
    +∑ * ⍺ (i, j) ⍺ (i, j) in {0,1}
    

    View Slide

98. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    27
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Pmap (i)
    Min.
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    { Ncycles / f (j) }
    △Texe (i) =Tmemory
    +∑ * ⍺ (i, j)
    Pmap (i) = Freq2Power (j) * ⍺ (i, j)
    ∑
    

    View Slide

99. Each Event:
    Objective:
    Constraints:
    Problem Formulation
    27
    ▸ Scheduling Problem → Constrained Optimization.
    N
    ∑
    i
    △Texe (i) x Pmap (i)
    Min.
    Order: ≤
    Tend (i) Tstart (i+1)
    Deadline: ≤
    Tstart (i) △Texe (i) TQoS (i)
    +
    { Ncycles / f (j) }
    △Texe (i) =Tmemory
    +∑ * ⍺ (i, j)
    Pmap (i) = Freq2Power (j) * ⍺ (i, j)
    ∑
    1
    =
    ⍺ (i, j)
    With the constraints: ∑
    

    View Slide

100. Each Event:
     Objective:
     Constraints:
     Problem Formulation
     27
     ▸ Scheduling Problem → Constrained Optimization.
     N
     ∑
     i
     △Texe (i) x Pmap (i)
     Min.
     Order: ≤
     Tend (i) Tstart (i+1)
     Deadline: ≤
     Tstart (i) △Texe (i) TQoS (i)
     +
     { Ncycles / f (j) }
     △Texe (i) =Tmemory
     +∑ * ⍺ (i, j)
     Pmap (i) = Freq2Power (j) * ⍺ (i, j)
     ∑
     1
     =
     ⍺ (i, j)
     With the constraints: ∑
     Only one setting
     is chosen
     

     View Slide

101. Each Event:
     Objective:
     Constraints:
     Problem Formulation
     27
     ▸ Scheduling Problem → Constrained Optimization.
     N
     ∑
     i
     △Texe (i) x Pmap (i)
     Min.
     Order: ≤
     Tend (i) Tstart (i+1)
     Deadline: ≤
     Tstart (i) △Texe (i) TQoS (i)
     +
     { Ncycles / f (j) }
     △Texe (i) =Tmemory
     +∑ * ⍺ (i, j)
     Pmap (i) = Freq2Power (j) * ⍺ (i, j)
     ∑
     1
     =
     ⍺ (i, j)
     With the constraints: ∑
     

     View Slide

102. Each Event:
     Objective:
     Constraints:
     Problem Formulation
     27
     ▸ Scheduling Problem → Constrained Optimization.
     N
     ∑
     i
     △Texe (i) x Pmap (i)
     Min.
     Order: ≤
     Tend (i) Tstart (i+1)
     Deadline: ≤
     Tstart (i) △Texe (i) TQoS (i)
     +
     { Ncycles / f (j) }
     △Texe (i) =Tmemory
     +∑ * ⍺ (i, j)
     Integer Linear
     Programing!
     Pmap (i) = Freq2Power (j) * ⍺ (i, j)
     ∑
     1
     =
     ⍺ (i, j)
     With the constraints: ∑
     

     View Slide

103. Putting Things Together
     28
     Web Application
     Hardware Architecture
     

     View Slide

104. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     

     View Slide

105. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     Predictor
     Events
     

     View Slide

106. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     Predictor
     Events
     Scheduler
     Predictions
     

     View Slide

107. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     Predictor
     Events
     Speculative
     
     Schedules
     Scheduler
     Predictions
     

     View Slide

108. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     Predictor
     Controller
     Events
     Speculative
     
     Schedules
     Uncommitted
     
     Results
     Scheduler
     Predictions
     

     View Slide

109. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     Predictor
     Controller
     Events
     Speculative
     
     Schedules
     Uncommitted
     
     Results
     Scheduler
     Predictions
     

     View Slide

110. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     Predictor
     Controller
     Events
     Speculative
     
     Schedules
     Commit
     Uncommitted
     
     Results
     Scheduler
     Predictions
     

     View Slide

111. PES
     Putting Things Together
     28
     Web Application
     Hardware Architecture
     Predictor
     Controller
     Events
     Speculative
     
     Schedules
     Commit
     Uncommitted
     
     Results
     Recover
     Scheduler
     Predictions
     

     View Slide

112. Proactive Event Scheduler
     Proactive Event Scheduler
     29
     Result reduce energy consumption meanwhile improve QoS
     Prediction
     Web Program
     
     Analysis
     Machine
     
     Learning
     Scheduling
     Constraint Optimization
     

     View Slide

113. Proactive Event Scheduler
     Proactive Event Scheduler
     29
     Result reduce energy consumption meanwhile improve QoS
     Prediction
     Web Program
     
     Analysis
     Machine
     
     Learning
     Scheduling
     Constraint Optimization
     

     View Slide

114. Experimental Setup
     30
     

     View Slide

115. Experimental Setup
     30
     Implemented our framework in
     Chromium on top of Android system
     

     View Slide

116. ODORID XU+E development board,
     which contains an Exynos 5410 SoC
     Experimental Setup
     30
     Implemented our framework in
     Chromium on top of Android system
     

     View Slide

117. ODORID XU+E development board,
     which contains an Exynos 5410 SoC
     Experimental Setup
     30
     Implemented our framework in
     Chromium on top of Android system
     UI-level record and replay for
     reproducibility. [ISPASS’15]
     

     View Slide

118. Evaluation
     ▸Baseline Mechanisms
     ▹Interactive governor (Interactive) — Android default
     ▹EBS: a state-of-the-art QoS-aware scheduler
     ▹Oracle: optimal scheduler
     31
     31
     Time
     Oracle
     input 1 input 2
     QoS
     
     Deadline 1
     QoS
     
     Deadline 2
     E1 E2
     E1 E2
     E1 E2
     QoS-Aware
     OS Governor
     

     View Slide

119. Evaluation
     ▸Baseline Mechanisms
     ▹Interactive governor (Interactive) — Android default
     ▹EBS: a state-of-the-art QoS-aware scheduler
     ▹Oracle: optimal scheduler
     31
     ▸Metrics
     ▹Energy Consumption
     ▹QoS Violation
     31
     

     View Slide

120. Evaluation
     ▸Baseline Mechanisms
     ▹Interactive governor (Interactive) — Android default
     ▹EBS: a state-of-the-art QoS-aware scheduler
     ▹Oracle: optimal scheduler
     31
     ▸Metrics
     ▹Energy Consumption
     ▹QoS Violation
     31
     ▸Applications
     ▹Top web applications (e.g., www.amazon.com)
     

     View Slide

121. Experimental Result
     32
     

     View Slide

122. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

123. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

124. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

125. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

126. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

127. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     QoS Violation
     0
     0.15
     0.3
     0.45
     0.6
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

128. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     QoS Violation
     0
     0.15
     0.3
     0.45
     0.6
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

129. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     QoS Violation
     0
     0.15
     0.3
     0.45
     0.6
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     

     View Slide

130. Experimental Result
     32
     Norn. Energy
     0
     0.25
     0.5
     0.75
     1
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     QoS Violation
     0
     0.15
     0.3
     0.45
     0.6
     163
     msn
     slashdot
     youtube
     google
     amazon
     ebay
     sina
     espn
     bbc
     cnn
     twitter
     Interactive EBS PES Oracle
     61% less QoS Violation and 26% of Energy Reduction
     

     View Slide

131. Conclusion
     33
     ‣ To better satisfy the user experience while
     minimizing the energy, coordinating across
     events is crucial for event-driven applications.
     

     View Slide

132. Conclusion
     33
     ‣ To better satisfy the user experience while
     minimizing the energy, coordinating across
     events is crucial for event-driven applications.
     ‣ PES combines statistic inference with
     application analysis on event prediction.
     

     View Slide

133. Conclusion
     33
     ‣ To better satisfy the user experience while
     minimizing the energy, coordinating across
     events is crucial for event-driven applications.
     ‣ PES achieves significant energy savings
     while reducing QoS violations.
     ‣ PES combines statistic inference with
     application analysis on event prediction.
     

     View Slide

134. PES can be applied to
     other event-driven applications!
     

     View Slide

135. Thanks
     

     View Slide