Power Management

1
Power Management
2
Power A New Resource

• Energy is consumed by
• Display
• WiFi card
• CPU
• Disk
• Memory Access

3
WiFi Card

• Operating Mode
• Send/Recv Mode (900-1400 mW)
• Sleep Mode (50mW)
• Proxy Assisted Power-Friendly Streaming to Mobile
  Devices
• P. Shenoy and P. Radkov, MMCN 2002

4
LCD Display

• Backlight modes
• Superbright (2.8W)
• Power Save (1.72W)
• Reducing Backlight Power Consumption for
  Streaming Video Applications on Mobile Handheld
  Devices
• S. Pasricha et. al. Workshop on Embedded Systems
  for Real-Time Multimedia, October 2003

5
CPU

• Running at
• 300MHz (22W)
• 700MHz (30W)
• 1GHz (40W)
• Practical Voltage Scaling for Mobile Multimedia
  Devices
• W. Yuan, K. Nahrstedt, ACM MM 2004

6
Power and Media Streaming

• Power Consumed When
• Receive packets
• Decode
• Display

7
Idea

• Send video stream that is power-friendly

8
Proxy-Assisted Power-Friendly Streaming to Mobile
Devices

• P. Shenoy and P. Radkov
• MMCN 2002

9
Architecture
Heres my energy budget for decoding network
reception and max resolution
server/proxy
client
10
Architecture
OK, what should I send?
server/proxy
client
11
Information Needed

• Map stream properties to energy requirement
• Need to know decoding time of a frame

12
Frame Decoding Time
13
Decoding Time

• As a function of size
• As a function of resolution

14
Estimating Frame Decoding Time
15
Transcoding

• If current stream would exceed client decoding
  energy budget
• Need to transcode by reducing quality

16
Transcode to what?

• E estimated energy needed
• while E gt energy budget
• reduce quality by e
• E estimate energy needed

17
Transcoded Streams
server/proxy
client
18
Estimating Frame Decoding Time
19
Reducing NIC Energy

• NIC has two modes active/sleep
• Client can activate NIC only when packets are
  expected.

20
Normal Transmission
21
Burst Transmission
I will start transmitting at 101254.86 pm
22
Transcoded Streams
OK, I will wake up at t kd
I will start transmitting at time t
server/proxy
client
23
Evaluation
Frame Number
24
Evaluation

• NIC Idle Uptime 2 20
• Energy Savings 65 98

25
Power and Media Streaming

• Power Consumed When
• Receive packets
• Decode
• Display

26
Reducing Backlight Power Consumption for
Streaming Video Applications on Mobile Handheld
Devices

• S. Pasricha et. al.
• ESTIMedia 2003

27
Idea

• If video is bright, can reduce backlight

28
Intensity of an Image
29
Perceived Intensity
30
Reduction in Backlight
31
Increment in Pixel Intensity
32
Compensate Every Frame?
33
How to compensate
playback at this backlight level
server/proxy
client
34
Evaluation

• Savings 350 650mW

35
Energy Efficient Soft Real-Time CPU Scheduling
for Mobile Multimedia Systems

• W. Yuan, K. Nahrstedt
• SOSP 2003

36
Motivation

• Mobile devices run on battery
• How to save battery?

37
Dynamic Voltage Scaling

• Example AMD Athlon 4 PowerOn 300, 500, 600,
  700, 800, 1000MHz
• Energy ? V2

38
CPU Scheduler

• When to execute a task
• How long to execute it
• How fast to execute it

39
Requirements

• I need C units of time, out of every T units.
• How to know C?

40
Probability Distribution
Cum. Prob.
cycles
41
Soft Real-Time Scheduling

• EDF with budget C
• If overrun budget, use available free time

42
CPU Requirements

• I need C units of time, out of every T units.

43
Speed Schedule
speed
time
44
Finding Speed Schedule

• Let task execute at speed vx during cycle x
• execution time
• energy per cycle
• expected energy

45
Optimize This

• Minimize
• Subjected to

46
Implementation

• Linux Kernel
• AMD Athlon 4
• 716 lines of code

47
Evaluation Cycles
48
Evaluation Power Savings
Stochastic Speed Schedule Missed Deadline EnergySavings
no no
yes no
yes yes