Software Controlled Power Management Yung-Hsiang Lu Giovanni De - - PowerPoint PPT Presentation

software controlled power management
SMART_READER_LITE
LIVE PREVIEW

Software Controlled Power Management Yung-Hsiang Lu Giovanni De - - PowerPoint PPT Presentation

Software Controlled Power Management Yung-Hsiang Lu Giovanni De Micheli Tajana Simunic Computer System Laboratory Stanford University Outline Dynamic Power Management (DPM) ACPI (Advanced Configuration & Power Interface)


slide-1
SLIDE 1

Software Controlled Power Management

Computer System Laboratory Stanford University Yung-Hsiang Lu Tajana Simunic Giovanni De Micheli

slide-2
SLIDE 2

Stanford University CODES 99 Rome, Italy 2

Outline

  • Dynamic Power Management (DPM)
  • ACPI (Advanced Configuration & Power Interface)
  • Software Architecture
  • Experimental Result
  • Conclusion
slide-3
SLIDE 3

Stanford University CODES 99 Rome, Italy 3

Power State Transitions (Hard Disk)

Working (spinning + IO) Sleeping (stop spinning) Idle (spinning)

read / write IO complete shut down read / write spin up

slide-4
SLIDE 4

Stanford University CODES 99 Rome, Italy 4

Dynamic Power Management

Shutdown a device when it is not used.

Workload Devices

requests power state commands

Power Manager

parameters request model computers embedded systems portable devices …..

Applications:

Implementation & evaluation Request and device modeling Management algorithms Power and performance tradeoffs

slide-5
SLIDE 5

Stanford University CODES 99 Rome, Italy 5

DPM is a Co-Design Issue

  • HW provides multiple power states

– allowing dynamic power state change – providing low state-transition overhead

  • SW implements DPM algorithms

– implementing adaptive, flexible, smart algorithms – using higher-level information

slide-6
SLIDE 6

Stanford University CODES 99 Rome, Italy 6

Our Goal

Build a framework to facilitate the design and evaluation of DPM algorithms on realistic workloads.

  • personal computer
  • commercial OS
  • real applications
slide-7
SLIDE 7

Stanford University CODES 99 Rome, Italy 7

OS OS

Kernel Kernel Power Management Power Management Device Device Driver Driver ACPI driver ACPI driver AML interpreter AML interpreter

ACPI Architecture

BIOS BIOS Applications Applications Platform Hardware Platform Hardware Motherboard Motherboard devices devices Chipset Chipset CPU CPU ACPI Tables ACPI Tables

ACPI ACPI

Table Table interface interface BIOS BIOS interface interface Register Register interface interface ACPI BIOS ACPI BIOS ACPI registers ACPI registers

slide-8
SLIDE 8

Stanford University CODES 99 Rome, Italy 8

Software Architecture

applications hardware devices device-specific driver power manager

  • perating system

requests power manager

ACPI

slide-9
SLIDE 9

Stanford University CODES 99 Rome, Italy 9

Our Software Architecture

applications hardware devices device-specific driver power manager

  • perating system

requests power manager command utilization Experiment DPM Algorithms power manager

ACPI

filter driver

slide-10
SLIDE 10

Stanford University CODES 99 Rome, Italy 10

Advantages

  • Separate policy and mechanism
  • Facilitate algorithm implementation &

evaluation

  • Achieve high-precision performance

analysis (microsecond)

slide-11
SLIDE 11

Stanford University CODES 99 Rome, Italy 11

Programming Interface

synchronization creating power command timing analysis

Power Manager Filter Driver

SetPowerState

  • bserving IO activities,

power state change…. Requests Arrive

slide-12
SLIDE 12

Stanford University CODES 99 Rome, Italy 12

Experiment Environment

  • Intel Pentium II-266 MHz
  • Windows NT 5 beta-2
  • 256 MB RAM
  • IBM Deskstar IDE HDD

– 3 power states – idle power 3.48 W – sleeping power 0.75 W – 7.2 sec / 52 J to wake up and 0.51 sec to sleep

slide-13
SLIDE 13

Stanford University CODES 99 Rome, Italy 13

Hard Disk State Transitions

working sleeping idle

read / write IO complete shut down read / write wake up timesleeping × Powersleeping + timeidle × Poweridle + # wakeups × Energywakeup Total Energy =

slide-14
SLIDE 14

Stanford University CODES 99 Rome, Italy 14

Power vs. Timeout

1.6 1.8 2 2.2 2.4 30 60 90 120 150 180 Timeout (sec) Power (W)

slide-15
SLIDE 15

Stanford University CODES 99 Rome, Italy 15

Power vs. Timeout

1.6 1.8 2 2.2 2.4 30 60 90 120 150 180 Timeout (sec) Power (W)

Adaptive timeout [Lu GLSVLSI 99]: 1.60 W

slide-16
SLIDE 16

Stanford University CODES 99 Rome, Italy 16

# shutdowns

30 60 90 120 30 60 90 120 150 180 Timeout (sec) # shutdown

slide-17
SLIDE 17

Stanford University CODES 99 Rome, Italy 17

% Waiting for Spinning Up

0% 2% 4% 6% 8% 10% 30 60 90 120 150 180 Timeout (sec) Waiting

slide-18
SLIDE 18

Stanford University CODES 99 Rome, Italy 18

Conclusion

  • DPM is a HW/SW collaborative

approach for reducing power.

  • We built a software architecture that

– is portable – facilitates DPM algorithm evaluation – runs realistic workloads

slide-19
SLIDE 19

Stanford University CODES 99 Rome, Italy 19

Acknowledgements

  • MARCO / SIA
  • Patrick Burke and Charles Orgish (Stanford)
  • Luca Benini and Alessandro Bogliolo (Bologna)