Low Power Design For Embedded Systems

1
Low Power Design For Embedded Systems
DYNAMIC POWER MANAGEMENT
tiánMiguel Ángel Sebastián González González
2
AGENDA

• INTRODUCTION
• What is Dynamic Power Managent?
• Power management.
• Break Even Time
• Power Management vs. Performance.

3
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

4
WHAT IS DYNAMIC POWER MANAGEMENT?

• Definition DPM is a design methodology for
  dynamically reconfiguring systems to provide the
  requested services and performance levels with a
  minimum number of active components or a minimum
  load on such components.

5
POWER MANAGEMENT

• System-level power management saves power of
  subsystems.
• If there were no overhead, power management would
  be trivial.
• a device should sleep only if the saved energy
  justifies the overhead.

6
BREAK -EVEN TIME

• Definition The minimum length of an idle period
  to save power.

7
(No Transcript)
8
POWER MANAGEMENT VS. PERFORMANCE

• Many policies focus on power saving and do not
  care the performance impact.
• Occasional delays are unavoidable .

9
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

10
PREDICTIVE TECHNIQUES

• Little knowledge of future input events.
• DPM based on uncertain predictions.
• Overprediction.
• Underprediction.

11
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

12
STATIC TECHNIQUES

• Fixed Timeout
• Predictive Shutdown
• Predictive Wakeup

13
FIXED TIMEOUT (I)

• Most common predictive PM policy
• Start a Timer at the beginning of an idle period
  TTO.
• If after TTO system still idle,PM forces Off
  state.

14
FIXED TIMEOUT (II)

• Advantages
• Generality.
• Safety improved increasing timeout values.
• Disadvantages
• Tradeoff efficiency for safety.
• waste power waiting for the timeout to expire.
• Pay a performance penalty upon wakeup.

15
PREDICTIVE SHUTDOWN

• Solve waste power
• 1 scheme Non linear regression equation
• If TPRED gt TBEgtshut down
• 2 scheme Based on a threshold.
• Tlt TTHR gt system is shut down.
• Not aplicable if scatter plot is not L-shaped.

16
PREDICTIVE WAKEUP

• Solve the performance penalty.
• Wakeup when the predicted idle time expires.
• Increase power dissipation if TIDLE has been
  underpredicted.
• Decreases delay for servicing the first incoming
  request after an idle period.

17
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

18
ADAPTIVE TECHNIQUES

• Static techniques ineffective when the workload
  is unknown or nonstationary.
• Several adaptive predictive techniques have been
  proposed
• Set of timeout and an index .
• Assigns a weight to timeouts.
• Only one timeout.

19
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

20
STOCHASTIC CONTROL

• Low power problem can be cast as a stochastic
  optimization problem.
• PM is modeled as a Markov process.
• complex systems with many power states.
• compute PM policies that are globally optimum
• explore tradeoffs between power and performance.

21
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

22
MODEL

• Generality
• High level of abstraction
• Non deterministic

23
MODEL
24
ELEMENTS OF THE MODEL (I)

• SERVICE REQUESTOR (SR)
• Send request to the SP.
• Modelled as a Markov chain
• Assume that the process and all its relevant
  parameters are know.

25
ELEMENTS OF THE MODEL (II)

• SERVICE PROVIDER (SP)
• Serves incoming requests from a workload source.
• Each state is characterized by a performance and
  power consumption level.

26
ELEMENTS OF THE MODEL (III)

• QUEUE
• When service requests arrive during one period,
  they are buffered in a queue.
• Several qeus disciplines, most common FIFO.
• POWERMANAGER
• communicates with the service provider and
  attempts to set its state at the beginning of
  each period.
• contains all proper specifications and collects
  all relevant information.
• Small power consumption

27
ELEMENTS OF THE MODEL (IV)

• DECISIONS
• PM observes the history of the system and
  controls the Service Provider by taking a
  decision.
• deterministic decision taking a single action on
  the basis of the history of the system.
• COST METRICS
• expected power consumption level c(sp,?s)
• performance penalty per unit time (depends on
  the queue length).

28
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

29
STATIC TECHNIQUES (I)

• The performance and power obtained by a policy
  are expected values.
• Policy optimization requires a Markov model for
  SP and SR.
• Can be safely assume that the SP model can be
  precharacterized, but not always the SR.

30
STATIC TECHNIQUE (II)

• Power consumption of the PM is not always
  negligible.
• Policy implementation may not be straightforward.
• Assumes a complete a priori knowledge of the
  system and its workload (SR) and workloads are
  often nonstationary

31
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

32
ADAPTIVE TECHNIQUES (I)

• based on three simple concepts
• policy precharacterization.
• parameter learning.
• policy interpolation.
• Two parameter Markov model for workload.
• Policy precharacterization constructs a
  two-dimensional (2-D) table.
• The table is filled by computing optimum policies
  under different workloads.

33
ADAPTIVE TECHNIQUES (II)

• Average techniques to estimate workload
  parameters based on past history.
• Parameter values are used to obatin the power
  management policy.
• If estimated parameter do not correspond to value
  in the table gt policy interpolation.

34
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

35
STOCHASTIC VS. PREDICTIVE TECHNIQUES

• Captures the global view of the system.
• It enables the exact solution.
• Exploits the strength and optimality of
  randomized policies.
• But more complicated.

36
AGENDA

• Dynamic Power Management Techniques.
• Predictive Techniques
• Static Techniques
• adaptive Techniques
• Stochastic Control
• Model
• Static Techniques
• adaptive Techniques
• Stochastic Vs.Predictive
• Conclusion

37
CONCLUSION

• DPM is a powerful design methodology for reducing
  power consumption
• Minimize power under performance constraints is
  the most important challenge.
• two groups of power management policies
  Predictives and Stochastic.
• Unexploted because of complexity of interfacing
  heterogeneous components.