A CPU Utilization Based Method for Dynamic Power Management

Zhu Jianquan

2007-01-01

Abstract:This paper suggest a dynamic algorithm whose parameter is adaptive according to the specific circumstances and the historic predicting results.The experimental result shows it is a big improvement in predicting CPU utilization.

What problem does this paper attempt to address?

A Self-aware Dynamic Power Management System

Chaoqiang Chu,De Ma,Xiaohua Luo

DOI: https://doi.org/10.3969/j.issn.1000-0755.2016.01.001

2016-01-01

Abstract:In this paper, we introduce a self-aware dynamic power management system, which is used to decrease the power consumption of the SoC system without performance cost by adjusting the working state of the SoC system dynamically. This dynamical working state adjustment is performed by the hardware automatically according to the feedback of the monitor module and the adjustment strategy configured by software in advance. This dynamic power management system is able to adjust the working state not only between different applications, but also between different stages in the same application if the occupancy of resources is changed. The adjustment is flexible and real-time.
Multi-core Chip Dynamic Power Management Framework Based on Reinforcement Learning br

Cheng Zhuo,Xudong Zeng,Yufei Chen,Songyu Sun,Guojie Luo,Qing He,Xunzhao Yin

DOI: https://doi.org/10.11999/JEIT220350

2023-01-01

Abstract:Multi-core chips can provide mighty computing capability for mobile intelligent terminals, but their performance is constraint by thermal and power issues. For this problem, this paper proposes a multi-core chipdynamic power management framework based on reinforcement learning. First, based on GEM5, a dynamicvoltage and frequency scaling simulation system of the multi-core chips is established. Second, a chip powermodel characterization method is adopted, which takes CMOS physical characteristics into consideration torealize online real-time power monitoring. Finally, a gradient reward method for the multi-core chips isdesigned, and a Deep Q Network (DQN) algorithm is used to learn the power management strategy for themulti-core chips. Compared with conventional Ondemand and MaxBIPS schemes, the simulation results showthat the proposed framework achieves 2.12% and 4.03% improvement in computational performance of the multi-core chips respectively
Analysis of Power Saving Effect for Dynamic Power Management

Wenzhi Chen,Huan Zheng

DOI: https://doi.org/10.1109/MESA.2006.296937

2007-01-01

Abstract:Dynamic power management (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. The principle of DPM selectively shuts down or depresses performance of some components that are idle. Many approaches have been exploited for DPM, different approach leads to different effects of power saving. Former analysis of power saving effect of DPM approach used to use theoretical computation or simulation which are difficult to reflect the real efficiency of power saving of the target DPM approach, this paper presents an analytic approach which could be embedded into the real implementation of DPM architecture to monitor the behavior of DPM approach, record information it needs for analysis, and analyze power saving effect offline, it could also support systems that have multi low power states. Experiment results shows that this analytic approach gets the same results as the simulated results
An Efficient and Flexible Learning Framework for Dynamic Power and Thermal Co-Management

Yuan Cao,Tianhao Shen,Li Zhang,Xunzhao Yin,Cheng Zhuo

DOI: https://doi.org/10.1145/3380446.3430640

2020-01-01

Abstract:At the era of Artificial Intelligence and Internet of Things (AIoT), battery-powered mobile devices are required to perform more sophisticated tasks featured with fast varying workloads and constrained power supply, demanding more efficient run-time power management. In this paper, we propose a deep reinforcement learning framework for dynamic power and thermal co-management. We build several machine learning models that incorporate the physical details for an ARM Cortex-A72, with on average 3% and 1% error for power and temperature predictions, respectively. We then build an efficient deep reinforcement learning control incorporating the machine learning models and facilitating the run-time dynamic voltage and frequency scaling (DVFS) strategy selection based on the predicted power, workloads and temperature. We evaluate our proposed framework, and compare the performance with existing management methods. The results suggest that our proposed framework can achieve 6.8% performance improvement compared with other alternatives.
Power Aware Job Scheduling with Quality of Service Guarantees : A Preliminary Study

Congfeng Jiang,Xianghua Xu,Jian Wan,Jilin Zhang,Yinghui Zhao

2010-01-01

Abstract:In large scale server systems, Quality of Service (QoS) performance cannot be guaranteed if only the objective is to minimize the total power consumptions separately. In this paper a power aware feedback control based power aware job scheduling algorithm (PFM) is proposed to minimize power consumption in computing system and to provide QoS guarantees. Simulations and experiments show that the proposed algorithm can reduce power consumptions significantly while still providing QoS guarantees and the performance degradation is acceptable. The results also show that conventional hardware based per-component and system-wide power management methods can save more power consumptions if they are in assistance with job-level adaptation.
User- and Process-Driven Dynamic Voltage and Frequency Scaling.

Bin Lin,Arindam Mallik,Peter Dinda,Gokhan Memik,Robert Dick

DOI: https://doi.org/10.1109/ispass.2009.4919634

2011-01-01

Abstract:We describe and evaluate two new, independently-applicable power reduction techniques for power management on processors that support dynamic voltage and frequency scaling (DVFS): user-driven frequency scaling (UDFS) and process-driven voltage scaling (PDVS). In PDVS, a CPU-customized profile is derived offline that encodes the minimum voltage needed to achieve stability at each combination of CPU frequency and temperature. On a typical processor, PDVS reduces the voltage below the worst-case minimum operating voltages given in datasheets. UDFS, on the other hand, dynamically adapts CPU frequency to the individual user and the workload through direct user feedback. Our UDFS algorithms dramatically reduce typical operating frequencies and voltages while maintaining performance at a satisfactory level for each user. We evaluate our techniques independently and together through user studies conducted on a Pentium M laptop running Windows applications. We measure the overall system power and temperature reduction achieved by our methods. Combining PDVS and the best UDFS scheme reduces measured system power by 49.9% (27.8% PDVS, 22.1% UDFS), averaged across all our users and applications, compared to Windows XP DVFS. The average temperature of the CPU is decreased by 13.2degC. User trace-driven simulation to evaluate the CPU only indicates average CPU dynamic power savings of 57.3% (32.4% PDVS, 24.9% UDFS), with a maximum reduction of 83.4%. In a multitasking environment, the same UDFS+PDVS technique reduces the CPU dynamic power by 75.7% on average.
Predictive Shutdown Method of Dynamic Power Management Based on Active Period

ZHAO Tan,CHEN Yu

DOI: https://doi.org/10.3969/j.issn.1001-3695.2007.07.018

2007-01-01

Abstract:A predictive shutdown method based on active period was proposed.This method utilized the relationship between active period and idle period,and also imported the dynamic adaptive factor.Not only does it improve the prediction accuracy of idle periods which change greatly,but also adjusts and adapts to the change in workload quickly.Experiments show that this method achieves better performance and energy saving than traditional ones.
Research on Processor Energy Saving Strategy Based on BP Neural Network

Bing GUO,Pengbo ZHANG,Yan SHEN,Yichun HUANG,Yabo CAO

DOI: https://doi.org/10.15961/j.jsuese.201700175

2018-01-01

Abstract:Focusing on dynamic power consumption of the chip,this paper proposed a new energy-saving strategy named BP-DVFS. Currently,DVFS (dynamic voltage and frequency scaling) technology failed to predict future behavior of CPU.In order to improve the perform-ance of CPU utilization and predict more accurate CPU dynamic frequency,it is essential to reduce its operating power consumption.A CPU mod-el named FPU-CPU (forward predict utilization) was built in strategy.A non-linear relationship between the CPU utilization in next period and the CPU running resource was assumed in the model.Five events related to CPU resources and extracted from run-time processor environment were measured as features.Afterwards,the training phase of was performed by the BP (back propagation) neural network.Various experiments were then carried out to predict different kinds of tasks including CPU utilization,processing and power consumption through BP neural network.To evalu-ate the performance of proposed method,the experimental results under same situation were compared with three traditional CPU frequency mod-ulation strategies.The experimental results showed that the CPU power consumption of BP-DVFS based strategy is lower than that of other three strategies when CPU handles different kinds of program tasks.The experimental verification indicated that the proposed technique wasnot only ac-curate inpredicting performance of CPU utilization,but also helpful to reduce power consumption of CPU.
Energy Efficiency Optimization Scheme for Multi-core ProcessorUnder Power Consumption Constraint

Xiaojie QIU,Hong AN,Junshi CHEN,Mengxian CHI,Xu JIN

DOI: https://doi.org/10.3969/j.issn.1000-3428.2017.04.007

2017-01-01

Abstract:Keeping processor power consumption under budget can reduce the cost of cooling and improve the system reliability.Most existing energy efficiency optimization schemes are profile-based offline schemes,which may reduce practicality.Furthermore,centralized algorithms which exhaustively search for the optimal configuration may be too complex.In this paper,a power consumption optimization scheme PPCM is proposed,which utilizes Dynamic Voltage and Frequency Scaling(DVFS) to control CPU power consumption under buidget and improve energy efficiency.PPCM decouples power consumption control and power consumption allocation to improve flexibility.It uses a dynamic linear model to estimate power consumption and manages it based on feedback control technology.It uses the ratio of computation to memory access as an indicator to allocate power consumption among applications,and then considers the features of multithread application and allocates power consumption among threads.Experimental results show that PPCM outperforms Priority algorithm by 10.7% in speed in average,5.1% in energy saving in average and Energy-Delay Product(EDP) is reduced by 14.3% in average.It is superior to PCMCA by 4.5% in speed in average and 5.0% in EDP in average.
Q-DPM: an Efficient Model-Free Dynamic Power Management Technique

M Li,XB Wu,R Yao,XL Yan

DOI: https://doi.org/10.1109/date.2005.247

2007-01-01

Abstract:When applying Dynamic Power Management (DPM) technique to pervasively deployed embedded systems, the technique needs to be very efficient so that it is feasible to implement the technique on low end processor and tight-budget memory. Furthermore, it should have the capability to track time varying behavior rapidly because the time varying is an inherent characteristic of real world system. Existing methods, which are usually model-based, may not satisfy the aforementioned requirements. In this paper, we propose a model-free DPM technique based on Q-Learning. Q-DPM is much more efficient because it removes the overhead of parameter estimator and mode-switch controller. Furthermore, its policy optimization is performed via consecutive online trialing, which also leads to very rapid response to time varying behavior.
Dynamic Power Management Policy Using Genetic Algorithm to Predict Idle Periods

KONG Fei,TAO Pin,YANG Shi-qiang

DOI: https://doi.org/10.3321/j.issn:1002-8331.2006.29.007

2006-01-01

Abstract:Reducing energy consumption has become one of the most important challenges in designing computing sys- tems.Dynamic power management policies exploit components’idle periods to save energy.If one idle period of some component is long enough, the component can be put into low power state during this period in order to reduce energy consumption.Many dynamic power management policies are based on predicting lengths of components’future idle peri- ods.The more accurate the prediction is, the more efficient the policy is.This paper proposes a novel idea of using ge- netic algorithm to predict lengths of future idle periods.We take K pairs of adjacent idle periods and active periods as a load- gene and define a kind of relationship between adjacent load- genes, and use genetic algorithm to predict future load- genes that most accords with the relationship, then the future idle periods’lengths can be computed with the pre- dicted load- genes.Experimental results show that the proposed policy has more accurate prediction in comparison with exponential- average approach, thus this policy can be used for dynamic power management more efficiently.
Robust and adaptive dynamic power management for time varying system

Min Li,Xiaobo Wu,Menglian Zhao,Ping Li,Xiaolang Yan

DOI: https://doi.org/10.1007/11535409_65

2005-01-01

Abstract:Dynamic Power Management (DPM) is an effective power reduction technique to dynamically control power state of system components. Although there are already a lot of papers on DPM, few of them present robust and adaptive solution to handle the inherent time varying behavior of real world system. In this paper, we propose techniques for targeting DPM on the time varying behaviors. In our approach, we apply efficient policy optimization method to generate online power management policy to achieve adaptiveness. Moreover, in order to handle the small scale varying behavior (perturbation) that tends to significantly degrade the performance of DPM, we apply the Bounded Parameter Markov Decision Process and Interval Value Iteration to derive robust policy tolerant to perturbations. Simulation results show that the proposed techniques are effective.
The Dynamic Power Management for Embedded System with Poisson Process

Jiangwei Huang,Tianzhou Chen,Minjiao Ye,Yi Lian

DOI: https://doi.org/10.1007/11535409_67

2005-01-01

Journal of Zhejiang University SCIENCE A

Abstract:The mass of the embedded systems are driven by second batteries, not by wired power supply. So saving energy is one of the main design goals for embedded system. In this paper we present a new technique for modelling and solving the dynamic power management (DPM) problem for embedded systems with complex behavioural characteristics. First we model a power-managed embedded computing system as a controllable Flow Chart. Then we use the Poisson process for optimisation, and give the power management algorithm by the help of Dynamic Voltage Scaling (DVS) technology. At last we built the experimental model using the PXA 255 Processors. The experimental results showed that the proposed technique can achieve more than 12% power saving compared to other existing DPM techniques.
Dynamic Power Management and Dynamic Voltage Scaling in Real-Time Cmp Systems

Lei Miao,Yong Qi,Di Hou,Chang-li Wu,Yue-hua Dai

DOI: https://doi.org/10.1109/nas.2007.23

2007-01-01

Abstract:Real-time embedded systems need reducing power consumption while still maintaining high performance. Dynamic power management (DPM) and dynamic voltage scaling (DVS) are effective control techniques to reduce the energy consumption. We present a real-time energy saving scheduling (RESS) algorithm using DPM and DVS to reduce the energy cost for chip multiprocessor (CMP) systems. The simulating experiment results indicate that our algorithm can save system energy cost with little system performance degradation.
Genetic Algorithm Based Idle Length Prediction Scheme for Dynamic Power Management

Fei Kong,Pin Tao,ShiQiang Yang,XiaoLi Zhao

DOI: https://doi.org/10.1109/cesa.2006.313542

2006-01-01

Abstract:Reducing energy consumption has become one of the most important challenges in designing computing systems. Dynamic power management policies exploit components' idle periods to save energy. If one idle period of some component is long enough, the component can be put into low power state during this period in order to reduce energy consumption. Many dynamic power management policies are based on predicting lengths of components' future idle periods. The more accurate the prediction is, the more efficient the policy is. This paper proposes a novel idea of using genetic algorithm to predict lengths of future idle periods. We take K adjacent idle periods and active periods as a load-gene and define some kinds of relationships between adjacent load-genes, then use genetic algorithm to predict future load-genes that most accords with the relationships. Experimental results show that the proposed scheme is more efficient than the exponential-average approach.
A New Dynamic Frequency Scaling Algorithm For Power-Aware And Real Time System

Chen Tianzhou,Qian Jie,Shi Qingsong,Huang Jiangwei

2006-01-01

Abstract:Scalability of the core frequency is a common feature of low-power processor architectures. Many methods for frequency scaling have been proposed to find the best trade-off between energy efficiency and computational performance. Embedded hardware monitors in the form of event counters have proven to offer valuable information in the field of performance analysis. We can get some clues from that, this will be helpful for us to find the target frequency. In this paper, we introduce a new algorithm to obtain the target frequency which benefit from the performance monitor counters. And once the system information changes, which means we can change the core frequency to perform at a lower power consumption, the core frequency will be changed through our method in the next time slice. Experimental results, based on measurements on the platform with Intel pxa255 processor, show that significant energy savings are achieved with little performance degradation. With some good situations the energy savings can go up to about 10% (with little performance loss). The proposed algorithm is effective in the power-aware and real time system.
Using dynamic and Static Approach with Compiler-Assisted in Power management for DVS-Enabled Embedded System

Tianzhou Chen,JiangWei Huang,Jie Qian,Liang Xiao

2007-01-01

WSEAS TRANSACTIONS ON COMPUTERS

Abstract:Most of embedded systems are drove by the battery. The capacity of the battery is limited. The high power consumption is becoming a critical problem for devices which are battery-powered. In this paper we presented a new approach for power management. When the program is compiling, we partitioned the Programs into different regions by Complier statically. And partitioning or merging regions dynamically according the performance statistic. The code of each region could be run on different CPU setting according to their charac- teristic. The different CPU setting could consume different power. We use the DVS technique to set state of CPU. The results showed that using dynamic and static approach together can save more energy than only using static method.
Power Consumption Prediction in Real-Time Multitasking Systems

Ernest Antolak,Andrzej Pułka

DOI: https://doi.org/10.3390/electronics13071347

IF: 2.9

2024-04-04

Electronics

Abstract:The paper presents a practical methodology for the prediction of power consumption of a multitask real-time system. The authors briefly recall the structure of the original multitasking time-predictable system. The system has a regular and scalable architecture. It consists of many cores based on pipeline processing with threads' interleaving mechanism. In previous works, the authors have presented a number of issues related to the methodology of designing a time-predictable system and the scheduling of hardware threads with different design goals and constraints. The current article presents an accurate method for measuring the dynamic power dissipated in a system for various scenarios and the system configuration. The proposed approach is based on continuous monitoring of the power consumed by the FPGA, transmitting the measurements to the computer from the measuring device, and online analysis of the obtained results. The authors based their power estimation method on the original task frequency factor (TF). A theoretical analysis was made, and a series of practical experiments were carried out. The results obtained enable accurate estimation of the power requirements for the tasks performed in the real-time system. The presented technique can be useful during the design stage of multitasking real-time systems.

engineering, electrical & electronic,computer science, information systems,physics, applied
Dynamic balancing method of power distribution and consumption tasks based on state iterative prediction and resource peaking shifting

Tiantian Cai,Junjian Chen,Junye Li,Ming Hu,Xiaohua Li,Zexiang Cai,Xuhui Wang

DOI: https://doi.org/10.3389/fenrg.2024.1398647

IF: 3.4

2024-05-04

Frontiers in Energy Research

Abstract:In the context of the new power system, the widespread access to massive distributed new energy sources has led to the power distribution and consumption tasks characterized by multiple time scales, wide random distribution, and large demand differences, resulting in unpredictable resource peaks in the tasks computing resource demand curve. In view of this situation, a method of forecasting and dynamic balancing of computing resource demand for power distribution and consumption tasks based on state iteration was proposed: firstly, the tasks computing resource demand model was established under the analysis of the attributes and parameter demand of the power distribution and consumption tasks scenario. Secondly, on the basis of the short-term effectiveness prediction of the traditional Markov model, the first-order difference of the state is used for data training to track the state fluctuation, and the historical state and the predicted state are used for state iteration, so as to avoid the convergence of long-term prediction. Finally, a dynamic balancing model is established according to the time-scale characteristics of cyclical and non-cyclical tasks, and the optimal configuration of load imbalance is achieved through the identification and adjustment of historical data and burst data. The simulation results show that the improved Markov model based on first-order difference and state iteration has the short-term accuracy of the traditional model and the long-term traceability of data fluctuations. The dynamic balancing model can combine the characteristics of historical data and burst data to effectively reduce the imbalance of resource demand, and show good ability to cope with resource imbalance deviation.

energy & fuels
Dynamic Software Allocation Algorithm for Saving Power in Pervasive Computing

Han Songqiao,Zhang Shensheng,Zhang Yong,Cao Jian

DOI: https://doi.org/10.3969/j.issn.1003-7985.2007.02.013

2007-01-01

Abstract:A novel dynamic software allocation algorithm suitable for pervasive computing environments is proposed to minimize power consumption of mobile devices. Considering the power cost incurred by the computation, communication and migration of software components, a power consumption model of component assignments between a mobile device and a server is set up. Also, the mobility of components and the mobility relationships between components are taken into account in software allocation. By using network flow theory,the optimization problem of power conservation is transformed into the optimal bipartition problem of a flow network which can be partitioned by the max-flow min-cut algorithm. Simulation results show that the proposed algorithm can save significantly more energy than existing algorithms.

A CPU Utilization Based Method for Dynamic Power Management

A Self-aware Dynamic Power Management System

Multi-core Chip Dynamic Power Management Framework Based on Reinforcement Learning br

Analysis of Power Saving Effect for Dynamic Power Management

An Efficient and Flexible Learning Framework for Dynamic Power and Thermal Co-Management

Power Aware Job Scheduling with Quality of Service Guarantees : A Preliminary Study

User- and Process-Driven Dynamic Voltage and Frequency Scaling.

Predictive Shutdown Method of Dynamic Power Management Based on Active Period

Research on Processor Energy Saving Strategy Based on BP Neural Network

Energy Efficiency Optimization Scheme for Multi-core ProcessorUnder Power Consumption Constraint

Q-DPM: an Efficient Model-Free Dynamic Power Management Technique

Dynamic Power Management Policy Using Genetic Algorithm to Predict Idle Periods

Robust and adaptive dynamic power management for time varying system

The Dynamic Power Management for Embedded System with Poisson Process

Dynamic Power Management and Dynamic Voltage Scaling in Real-Time Cmp Systems

Genetic Algorithm Based Idle Length Prediction Scheme for Dynamic Power Management

A New Dynamic Frequency Scaling Algorithm For Power-Aware And Real Time System

Using dynamic and Static Approach with Compiler-Assisted in Power management for DVS-Enabled Embedded System

Power Consumption Prediction in Real-Time Multitasking Systems

Dynamic balancing method of power distribution and consumption tasks based on state iterative prediction and resource peaking shifting

Dynamic Software Allocation Algorithm for Saving Power in Pervasive Computing