Temperature-aware Operating System Scheduling (presentation)

Temperature-Aware Operating System Scheduling Eugene Otto Mar. 30, 2006 Technical Advisor: Prof. Kevin Skadron STS Advisor: Prof. Helen Benet-Goodman

Overview  Key

concepts  Problem/Solution  Social context  Testing/Results

Key Concept: Multitasking

Solution

Social Impact  Improves

reliability for systems operating in high temperature environments (Ex: Iraqi desert).

 Improves

reliability for e-commerce solutions (Ex: Stock trading).

 Reduces

fan noise (DVD playback).

Results of First Benchmark

Temp (°C) Time (s)

Max Min

Fan Speed (RPMs)

Avg.

Max

Min

Avg.

Original, Unthrottled

359

80

51

72.78

5940

3780 4958.947

Original, Throttled

6749

56

53

55.11

3840

3780 3808.684

Modified, Throttled

1885

56

48

54.41

4080

4020 4079.842

Temperature (°C)

Change in Temperature as Frames are Decoded 90 80 70 60 50 40 30 20 10 0

Original, Unthrottled

Original, Throttled Modified, Throttled

0

10000

20000

30000

Frames

40000

50000

Fan Speed (RPMs)

Change in Fan Speed as Frames are Decoded 7000 6000 5000 4000 3000 2000 1000 0

Original, Unthrottled

Modified, Throttled

0

10000

20000

Original, Throttled

30000

Frames

40000

50000

Results of Second Benchmark Throttling Level

Original Scheduler

Modified Scheduler

0.0%

Yes

-

12.5%

Yes

-

25.0%

Yes

-

37.5%

No

Yes

50.0%

-

Yes

62.5%

-

No

75%

-

-

87.5%

-

-

Conclusion  Modified

scheduler plays DVD video at lower temperatures with less fan noise.

 Paves

way for research into more sophisticated temperature-aware scheduling algorithms.

Questions?