Best detection scheme achieves 100% hit detection with <5% false alarms

1
Runtime Power Monitoring and Phase Analysis
Methods for Power Management
Canturk Isci and Margaret Martonosi
Princeton University
Motivation and Research Overview
Live, Runtime Power Monitoring and Estimation

• Power is the primary design constraint for
  current systems
• Power density ? Cooling / Thermal constraints
• Energy ? Battery life
• Workloads exhibit drastically different behavior
  both within applications and among different
  applications (Phases)
• These can be exploited by workload directed
  dynamic management techniques
• Dynamically reconfigurable hardware
• Power balancing / Activity migration
• Need methods to track application power behavior
  and identify different (repetitive) regions of
  operation
• Live, real-system experiments
• Reflect behavior of real, modern processors
• Observe long time periods
• Guide on-the-fly adaptations

Our Work

• Counter Based Power Estimation
• Idealized view For all components on a chip.

Total Power Estimates and Measurement Validation
Power of component I
? Monitor application Execution -
Performance behavior via performance
monitoring counters (PMCs) - Control flow via
dynamic instrumentation
MaxPowerI ArchScalingI AccessRateI
From Microarch. Properties
Die Area Stressmarks
CPU Performance Counters!
? Represent application execution as a stream
of PMC and control flow samples ? Estimate
power behavior from PMC information ? Apply
phase tracking, detection and prediction
strategies under real-system effects based on PMC
and control flow features

• Realistic view Handle non-linear scaling

Empirical Multimeter Measurements
NonGatedPowerI
? Use application phase information to guide
dynamic/adaptive power management techniques
Per-Component Estimates Ex. Equake
Fast (Real-time) Offers estimated view of
on-chip detail for real systems Real
measurement validation

• Initialization and computation phases
• Initialization with high complex IA32
  instructions
• FP intensive mesh computation phase

? Employ real power measurements to provide
feedback to runtime power estimations and to
evaluate phase characterizations
Applications of Power Phase Analysis
Power Phase Analysis on Real Systems

• Phases Distinct and often-recurring regions of
  program behavior
• Ex Vortex

• Phase Detection Under Real-System Variability
• Problem Definition Variability effects on phases

• Long-Term Value and Duration Prediction of Memory
  Bound Phases for DVFS

• Evaluating Control-Flow-Based and Event-Counter
  Based Approaches

Control flow (Basic Block Vectors / BBVs)
A B C B
Ideal
Glitch
A B C B D B
Gradient
A B C B D E B
Shift
A B C B D E B
Event counters (PMCs)
Mutation
A B C B D E F
Time Dilation
A B C B D E F

• Proposed Solution Transition-guided phase
  detection framework

• Experimentation

• Can predict gt90 of DVFSable phases, with less
  than 5 prediction overshoots!

• Power can also exhibit phase behavior

1
1
1
000
000
000
000

• Power Balancing for Multiprocessor Systems /
  Activity Migration

A B C B
t
Power
Power
Task1
Task2
1
1
1
1
1
1
000
000
000
000
000
t
Swap hot task
A B C B D E F
Core/µP 1
Core/µP 2

• Mutations ? Transition based tracking
• Glitches and gradients ? Glitch/Gradient
  Filtering
• Shifts ? Binary cross correlations
• Time Dilations ? Near-neighbor blurring

• Phase Tracking By evaluating the similarity
  among PMC vectors (PVs)
• Similarity Criterion L1-Distance between PVs

Speed up!
Slow down!
run1

• Evaluation

Match!
1
Conclusions
t

• Both approaches bring significant insights to
  application power behavior

run2

• PVs achieve lt 5W within phase variations with lt10
  phases

• Certain compositions of event counters can
  provide reasonably accurate runtime estimates for
  processor power consumption and distribution of
  power among architectural components
• Workloads exhibit phases in their performance as
  well as power behavior- Performance counter
  vectors help identify different (recurring) power
  phases of applications
• Real system variability effects impose additional
  challenges for detecting recurrent phases- Phase
  transition guided approach, together with
  supporting methods such as glitch/gradient
  filtering and near-neighbor blurring enable
  detection of repetitive power phase behavior
• Both control flow and event counter based
  application features provide insight to
  application power behavior- PMC based approaches
  generally provide a better proxy to application
  power phase behavior, due to their strong
  physical binding to processor power consumption
• These phase oriented methods can be employed to
  guide range of applications in current and next
  generation systems

t

• Real-System Effects on Phases Metric and time
  variability

0 detect threshold?Phit 1Pfalse alarm 1
Best detection scheme achieves 100 hit
detection with lt5 false alarms

• PMCs achieve (on average 40) less errors than
  BBVs in power phase characterization

Very high detect threshold?Phit 0Pfalse
alarm 0