Power Management for Mobile Operating System

Embed Size (px)

Citation preview

  • 8/2/2019 Power Management for Mobile Operating System

    1/22

    Hans Joshep V. OBrien

    Lanie M. Arco

    Power Management for MobileOperating System

  • 8/2/2019 Power Management for Mobile Operating System

    2/22

    Introduction Batteries

    Reducing physical dimension

    If developers successfully reduce the basic

    consumption of mobile computing devices, usersmay have the bonus of either retaining the chargecharacteristics of the battery even the batterydimension is reduced and also increase thecharge time of the battery.

  • 8/2/2019 Power Management for Mobile Operating System

    3/22

    Introduction There are components in mobile computing

    devices that are responsible for powerconsumption, these includes the physicalcomponents.

  • 8/2/2019 Power Management for Mobile Operating System

    4/22

    Power Consumption of CPU CPUs clock rate

    is related to the power consumed by the CPU, thesupply voltage, and the capacitance of the devicesbeing switched (e.g. the transistors).

    - when the CPUs clock rate decreases the powerconsumption also deceases, clock rate decreasesbecause of the switching characteristics of the logicgates (Gregory F. Welch, 1994).

  • 8/2/2019 Power Management for Mobile Operating System

    5/22

    Scheduling for Reduced CPU Energy

    CPUs power consumption is significant in mobilecomputing device.

    If the CPU clock frequency and supply voltagecan be controlled, linear and quadratic savings inpower can be realized.

    Because lowering the supply voltage results in aquadratic power savings, this is the authors

    preferred method for power reduction.

  • 8/2/2019 Power Management for Mobile Operating System

    6/22

    Scheduling for Reduced CPU Energy

    However, the supply voltage cannot be lowered(dynamically) without also reducing the clockspeed.

    The authors consider the CPU clock rate to belinearly adjusted with the supply voltage.

  • 8/2/2019 Power Management for Mobile Operating System

    7/22

    Method Three algorithms for adjusting the CPU clock

    speed under the control of the operating system.

    The evaluation of the algorithms wasaccomplished through trace-driven simulations.

  • 8/2/2019 Power Management for Mobile Operating System

    8/22

    Results The three algorithms basic approach is to

    balance the usage between the periodic bursts ofhigh CPU utilization and the remaining idle time ofthe CPU.

    For instance, lets assume that we have a job that

    has a CPU burst time of 10 milliseconds usage atthe full CPU clock speed and followed by another10 milliseconds idle time.

  • 8/2/2019 Power Management for Mobile Operating System

    9/22

    Results If it is possible that that job can be replaced by a

    20 milliseconds period of CPU usage at half thenormal clock speed, without affecting the useradversely, then such a scheme can be used to

    conserve power (Welch, G. 1994).

  • 8/2/2019 Power Management for Mobile Operating System

    10/22

    Results All three algorithms rely on the assumption that

    the operating system sleepevents that normallyresult in CPU idle time and can be classified intotwo categories:

    - hard and soft sleep events.

  • 8/2/2019 Power Management for Mobile Operating System

    11/22

    Results Hard sleep events

    events result in idle time during which the CPUspeed cannot be reduced.

    Soft sleep events

    - result in idle time during which the CPUspeed can be reduced

  • 8/2/2019 Power Management for Mobile Operating System

    12/22

    Results The authors classify of CPU idle time because

    they want to be fair about assessments, the

    authors want to know which idle times weresubject for use in balancing periods of high and

    low activity.

  • 8/2/2019 Power Management for Mobile Operating System

    13/22

    Discussions Three algorithms by Weiser et al. 1994

    OPT

    FUTURE

    PAST

  • 8/2/2019 Power Management for Mobile Operating System

    14/22

    Discussions

    OPT Algorithm is an impractical and undesirable algorithm. It is

    completely optimistic (and impractical) in the sensethat it assumes complete knowledge of the future

    work to be done in an interval (Welch, G. 1994).The said algorithm is undesirable because itrigorously affect the run times of user tasks anddisregarding importance of events such as

    keystroke response or network communications.

  • 8/2/2019 Power Management for Mobile Operating System

    15/22

    Discussions

    FUTURE Algorithm- is similar to OPT because it also assumes future

    task to be done but it does so for only shortwindows of time, and optimizes over only those

    windows.

  • 8/2/2019 Power Management for Mobile Operating System

    16/22

    Discussions FUTURE is unrealistic because it peers into the

    future.

    Unlike OPT, it is practical because it onlyoptimizes over short windows, therefore the effect

    and impact on time-critical work is minimized.

  • 8/2/2019 Power Management for Mobile Operating System

    17/22

    Discussions PAST

    Unlike OPT and FUTURE algorithms PASTalgorithm is a realistic (practical) version of thesealgorithms.

    PAST algorithm is realistic because instead ofpeering into the future, it looks to a short window inthe past for information about CPU usage.

  • 8/2/2019 Power Management for Mobile Operating System

    18/22

    Discussions

    - It is desirable because like FUTURE it onlymonitors activity over a short window of the trace.But of the disadvantage of this algorithm is that itassumes the activity in the next (future) window

    will be like that in the current (past) window.

  • 8/2/2019 Power Management for Mobile Operating System

    19/22

    Discussions

    - it is indeed more profitable (in terms of powersavings) to spread out work when possible overperiods of slower CPU clock rates, rather thancontending with bursts of high-speed activity

    followed by wasted idle time.

  • 8/2/2019 Power Management for Mobile Operating System

    20/22

    Discussions

    - The authors also concluded that in theirsimulations a PAST implementation with a 50 mswindow would have saved up to 50% of thepower in a 3.3 volt CPU

  • 8/2/2019 Power Management for Mobile Operating System

    21/22

    Conclusion The development and rise of mobile computing

    has introduced new challenges and issues for thedesigners of computers and computer operatingsystems.

    These techniques together focus on what aretypically the main causes in terms of powerconsumption in mobile computing: CPU/memorydevices.

  • 8/2/2019 Power Management for Mobile Operating System

    22/22

    Conclusion While hardware advances have and will likely

    continues to reduce the power consumption ofthese devices, efficient operating systemtechniques significantly may reduce power

    consumption without considerably affecting theperceived performance.