Download presentation
Presentation is loading. Please wait.
Published byAmanda Logan Modified over 9 years ago
1
NUS.SOC.CS5248 A Time Series-based Approach for Power Management in Mobile Processors and Disks X. Liu, P. Shenoy and W. Gong Presented by Dai Lu
2
NUS.SOC.CS5248 Contents Introduction Time Series based Power Management Utilization Measurement Prediction Model Speed Setting Strategy Implementation Evaluation Summary
3
NUS.SOC.CS5248 Introduction Multimedia applications prevalent on mobile devices 3G/4G wireless network Devices more and more powerful Samsung SPH-V5400 hand phone is equipped with a 1.5 GB micro drive Energy is a scarce resource
4
NUS.SOC.CS5248 Previous Work CPU DVFS: Dynamic Voltage and Frequency Scaling Infer task periodicity by work-tracking heuristic Assume implicit deadlines for interactive applications Only periodic applications; assumes applications tell OS their periods and work amount Disk DRPM: Dynamic Rotations Per Minute Monitor disk request queue length On-disk cache impact not considered
5
NUS.SOC.CS5248 Why DRPM? Power- RPM relation K e : spindle motor voltage R: motor resistance ω: angular velocity Similar to DVS for processors (P~fV 2 )
6
NUS.SOC.CS5248 Contents Introduction Time Series based Power Management Utilization Measurement Prediction Model Speed Setting Strategy Implementation Evaluation Summary
7
NUS.SOC.CS5248 New Work Low overhead Prediction with simple statistical model in time series analysis Processor + disk TS-DVFS + TS-DRPM Different CPU scaling factor for different tasks Enable coexistence of MM and non-MM applications
8
NUS.SOC.CS5248 TS-PM enabled OS kernel
9
NUS.SOC.CS5248 Prediction Model Box-Jenkins model in time series analysis Assume a stationary process Statistical properties (mean, variance) are essentially constant through time. Firs-order autoregressive process (AR(1)) predictor ũ t = Φ 1 ũ t-1 +a t Φ 1 : Correlation coefficient a t : Error/ random shock Sample Autocorrelation Function (SAC)
10
NUS.SOC.CS5248 Prediction Model Cont. Estimated demand: Estimated mean: Estimated constant( SAC): TS-DVFS: one AR(1) for every task TS-DRPM: a single AR(1)
11
NUS.SOC.CS5248 Measuring utilization CPU e: full-speed execution time q: time quantum allocated to the task Disk r: response time s: scaling factor
12
NUS.SOC.CS5248 Speed Setting Strategy TS-DVFS Two level CPU setting Interval T Subinterval within T
13
NUS.SOC.CS5248 Speed Setting Strategy TS-DRPM Performance slow-down P diff [i] = a(1-h) × T × R diff [i] Estimated utilization û i = û + P diff [i]/ T h: hit rate a: arrival rate R diff : rotational latency difference Choose the lowest RPM level satisfying (û i - û max ) / û max ≤ threshold
14
NUS.SOC.CS5248 Contents Introduction Time Series based Power Management Utilization Measurement Prediction Model Speed Setting Strategy Implementation Evaluation Summary
15
NUS.SOC.CS5248 Implementation CPU 300-677 MHz, Transmeta Divide into 5 steps Mapping scaling factor to frequency level Disk 3000-5400 RPM Divide into 5 steps Assumed power consumption level Trace driven simulation with DiskSim
16
NUS.SOC.CS5248 Frequency and RPM Mapping
17
NUS.SOC.CS5248 Contents Introduction Time Series based Power Management Utilization Measurement Prediction Model Speed Setting Strategy Implementation Evaluation Summary
18
NUS.SOC.CS5248 TS-DVFS Up to 38.6% energy saving against LongRun
19
NUS.SOC.CS5248 TS-DRPM Up to 20.3% saving against TPM perf (oracle)
20
NUS.SOC.CS5248 Summary Time series statistical model TS-DVFS TS-DRPM Comments General PM, no QoS measurement like deadline miss rate Multiple rotational speed disk not commercially available Increase the accuracy of profiling disk access patterns. “Hit if response time < τ, otherwise miss. ”
21
NUS.SOC.CS5248 References Chameleon: Application Controlled Power Management with Performance Isolation, X. Liu and P. Shenoy, Technical report 04-26, Department of Computer Science, University of Massachusetts Forecasting and time series: an applied approach 3 rd ed, Bowerman and O’Connell, Duxbury, 1993 Reducing disk power consumption in servers with DRPM, S. Gurumurthi, A. Sivasubramaniam and H. Franke, IEEE Computer, Dec 2003
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.