Dynamic Voltage Frequency Scaling Research Articles

Noise-Aware DVFS for Efficient Transitions on Battery-Powered IoT Devices

Low power system-on-chips (SoCs) are now at the heart of Internet-of-Things (IoT) devices, which are well-known for their bursty workloads and limited energy storage—usually in the form of tiny batteries. To ensure battery lifetime, dynamic voltage frequency scaling (DVFS) has become an essential technique in such SoC chips. With continuously decreasing supply level, noise margins in these devices are already being squeezed. During DVFS transition, large current that accompanies the clock speed transition runs into or out of clock networks in a few clock cycles, induces large ${\text {L}di}{/}{\mathrm {d}t}$ noise, thereby stressing the power delivery system (PDS). Due to the limited area and cost target, adding additional decoupling capacitance to mitigate such noise is usually challenging. A common approach is to gradually introduce/remove the additional clock cycles to increase/decrease the clock frequency in steps, also known as, clock skipping. However, such a technique may increase DVFS transition time, and still cannot guarantee minimal noise. In this paper, we propose a new noise-aware DVFS sequence optimization technique by formulating a mixed 0/1 programming to resolve the problems of clock skipping sequence optimization. Moreover, the method is also extended to schedule extensive wake-up activities on different clock domains for the same purpose. The experiments show that the optimized sequence is able to significantly mitigate noise within the desired transition time, thereby saving both power and energy.

Frequency regulation service provision in data center with computational flexibility

The rapid adoption of cloud storage and computing services led to unprecedented growth of data centers in the world. As bulk energy consumers, large-scale data centers in the U.S. rack up billions in electricity costs annually. Fortunately, the operational flexibility of data centers can be leveraged to provide valuable frequency regulation services in smart grids to mitigate the indeterminacy of the renewable generation resources. Specifically, this paper aims to leverage computational flexibility provided by servers, such as dynamic voltage frequency scaling and dummy loads. This paper develops a comprehensive framework for data center’s frequency regulation service provision in both hour-ahead market and real-time operations. A risk constrained hour-ahead bidding strategy along with a real-time data center power consumption control algorithm are developed to minimize electricity bills and the total response time of the requests. The introduction of dummy load, realistic bi-linear server power consumption model, and probabilistic forecast of electricity and frequency regulation service prices enable the data center to accurately follow frequency regulation signals, while reducing the financial risks associated with electricity market participation. The simulation results show that the proposed frequency regulation provision framework results not only in significant cost reduction for data centers, but also limits degradation in quality of service. Meanwhile, the stability and reliability of a power grid will be improved by the frequency regulation service provision.

BrownoutCon: A software system based on brownout and containers for energy-efficient cloud computing

VM consolidation and Dynamic Voltage Frequency Scaling approaches have been proved to be efficient to reduce energy consumption in cloud data centers. However, the existing approaches cannot function efficiently when the whole data center is overloaded. An approach called brownout has been proposed to solve the limitation, which dynamically deactivates or activates optional microservices or containers. In this paper, we propose a brownout-based software system for container-based clouds to handle overloads and reduce power consumption. We present its design and implementation based on Docker Swarm containers. The proposed system is integrated with existing Docker Swarm without the modification of their configurations. To demonstrate the potential of BrownoutCon software in offering energy-efficient services in brownout situation, we implemented several policies to manage containers and conducted experiments on French Grid’5000 cloud infrastructure. The results show the currently implemented policies in our software system can save about 10%–40% energy than the existing baselines while ensuring quality of services.

Reduced carbon emission and optimized power consumption technique using container over virtual machine

Environmental warning is caused by IT industry which critically leads the global pollution with huge amount of toxic carbon emission which is drastically increasing day by day due to demand and usage raised. Due to the environmental warning the industries are in awful need of reducing the carbon foot print by inducing green computing. This paper has achieved green computing by implementing two different algorithms (1) Water Shower Model (WSM) and (2) Trigger—WSM load balancing algorithms, and two different techniques (3) Recommending Containers over virtual machine techniques and 4. DVFS (Dynamic Voltage Frequency Scaling) modeling. The observation between the recommendation systems for container over virtual machine for a sample of four containers with one application each and four virtual machine with one application each is monitored for carbon emission equivalent in kg co2 for about 1 week is 14 kg co2 for container and 84.4 kg co2 for virtual machine, where a drastic difference in the amount of carbon emission is seen. So recommending container will be the best possible solution for the IT based on applications, by enforcing these ideas and techniques the carbon emission can be drastically decreased and the amount of carbon footprint in the atmosphere will also be reduced. The amount of power consumption utilized for the same model is 15.71367 W for container and 94.72667 W for virtual machine is also observed; in the field of IT the power consumption also to be taken into consideration for reducing carbon emission. The recommendation system along with the proposed algorithm will reduce the amount of carbon footprint in the environment.

Power Noise and Near-Field EMI of High-Current System-in-Package With VR Top and Bottom Placements

Integrated voltage regulators (IVRs) support an increasing number of voltage domains while enhancing the transient response and enabling fine-grained dynamic voltage frequency scaling. A voltage regulator (VR)-on-package is a promising IVR topology, which supports high-voltage transmission within a printed circuit board and package, leading to lower distribution loss. A power delivery network within a VR-on-package environment is presented in this paper. This paper also presents a study of VR top- and bottom-placement topologies within a system-in-package environment in terms of electromagnetic interference (EMI) and power integrity. This comparative analysis targets a specific server package application, and the power integrity evaluation is focused on the power delivery network between the VR and IC. The VR top-placement topology exhibits more than $3\times $ less EMI and 15.3% lower worse case IR drop as compared with the VR bottom-placement topology. The tradeoffs are, however, a greater package power loss and higher package cost due to the larger number of package layers in the VR top-placement topology. The VR top-placement topology exhibits a 52.6% higher power loss through the package than the VR bottom-placement topology.

A green SLA constrained scheduling algorithm for parallel/scientific applications in heterogeneous cluster systems

Cloud services utilize Heterogeneous Clusters as the hardware infrastructure at the backend. Minimal energy consumption in Heterogeneous Cluster System (HCS) is an important but challenging job. It can bring numerous benefits such as drop in operational cost, enhanced system reliability and green environment. Scheduling on HCS, with the objectives of minimizing the energy consumption, is being addressed by the research community at large. This work proposes a scheduling algorithm with green Service Level Agreement (SLA) constrained execution of parallel applications. The green SLA assumes that the tasks are to be scheduled on Dynamic Voltage Frequency Scaling (DVFS) enabled HCS. The proposed algorithm acts in two phases: first, it minimizes the energy consumption for the slacks of the schedule using DVFS technique, and second, a green SLA is made between a customer and service provider to maintain a good energy-performance trade-off. With this strategy, makespan of the schedule is extended a bit but the energy consumption is minimized for the period of extended makespan. It is observed that the decrease in one level of voltage and frequency of high-speed Processing Element (PE) saves more energy compared to that of low-speed PE. To utilize this, jobs running on higher voltage and frequency level of the PE are prioritized first. The performance of the proposed algorithm is studied by simulation on several standard Directed Acyclic Graphs (DAGs) and few other well-known real-world scientific workflows. The comparative study exhibits the effectiveness of the proposed algorithm over state of the art on four real heterogeneous PEs.

Energy Minimization for Multicore Platforms Through DVFS and VR Phase Scaling With Comprehensive Convex Model

Energy management is a critical challenge in multicore processors due to continuous technology scaling. Previous methods have mostly focused on the energy minimization of the processor cores. However, energy overhead of the off-chip voltage regulator (VR) has recently shown to be a nontrivial part of the total energy consumption and has been previously overlooked. In this paper, we propose an overall energy optimization method for the system that minimizes both per-core energy consumption and VR energy consumption using dynamic voltage frequency scaling and VR phase scaling by solving a comprehensive convex model. In order to improve the accuracy of the task latency model, a new task model considering both computation and memory access of the task is also developed. Furthermore, for better scalability and lower online overhead, we decompose our proposed convex method into two stages: 1) an offline stage and 2) an online stage. During the offline stage, we explore the convex model by assuming different numbers of active phases of the VR, various workload pressures and workload characteristics to collect the optimal frequency assignments under different scenarios. During the online stage, the specific frequency assignment for cores and optimal active phase number of the VR are selected and applied based on the actual workload pressure and its characteristics running on the cores. Experiments on real benchmarks show that when compared with the state-of-the-art approaches, which are oblivious to VR overheads and exploit slack time to achieve energy minimization, our method can achieve a significant energy saving of up to 22.4% with negligible online overhead.

Recent advances in cloud data centers toward fog data centers

Recent advances in cloud data centers toward fog data centers

Minimizing energy consumption with reliability goal on heterogeneous embedded systems

The embedded systems generally require to be low-powered and highly reliable. In order to achieve the low-power design goal, dynamic voltage frequency scaling (DVFS) technique has been widely employed in various embedded application scenarios. However, DVFS reduces execution frequency, which increases transient faults of the processor dramatically. As a result, the reliability of the application will be severely reduced. In this paper, we aim at minimizing energy consumption with reliability goal for parallel application on heterogeneous embedded systems. Since the reliability of the application is the product of the reliability of all the tasks that belong to the application, the reliability goal of the application is transformed into the reliability goal of each task. Considering that some systems may not support DVFS techniques, two methods are proposed to transform the reliability goal of the application into each task for non-DVFS and DVFS, respectively. Based on the reliability goal transformation methods, two energy-efficient scheduling algorithms with the reliability goal are designed. Experiments with real parallel applications demonstrate that the proposed algorithms have significant improvements in energy efficiency compared with the state-of-the-art algorithms.

Allocation of energy-efficient task in cloud using DVFS

Nowadays, the expanding computational capabilities of the cloud system rely on the minimisation of the consumed power to make them sustainable and economically productive. Power management of cloud data centres received a great attention from industry and academia as it consumes high energy and thus increases the operational cost. One of the core approaches for the conservation of energy in the cloud data centre is the task scheduling. This task allocation in a heterogeneous environment is a well known NP-hard problem due to which researchers pay attention for proposing various heuristic techniques for the problem. In this paper, a technique is proposed based on dynamic voltage frequency scaling (DVFS) for optimising the energy consumption in the cloud environment. The basic idea is to address the trade-off between energy consumption and makespan of the system. Here, we formally introduce a model that includes various subsystems and assess the implementation of the algorithm in the heterogeneous environment.

IBrownout: An Integrated Approach for Managing Energy and Brownout in Container-Based Clouds

Energy consumption of Cloud data centers has been a major concern of many researchers, and one of the reasons for huge energy consumption of Clouds lies in the inefficient utilization of computing resources. Besides energy consumption, another challenge of data centers is the unexpected loads, which leads to the overloads and performance degradation. Compared with VM consolidation and Dynamic Voltage Frequency Scaling that cannot function well when the whole data center is overloaded, brownout has shown to be a promising technique to handle both overloads and energy consumption through dynamically deactivating application optional components, which are also identified as containers/microservices. In this work, we propose an integrated approach to manage energy consumption and brownout in container-based cloud data centers. \color{black} We also evaluate our proposed scheduling policies with real traces in a prototype system. The results show that our approach reduces about 40%, 20% and 10% energy than the approach without power-saving techniques, brownout-overbooking approach and auto-scaling approach respectively while ensuring Quality of Service.

Fuzzy Decision Efficiency for Multiprocessor Architecture Configuration

Dynamic voltage/frequency scaling (DVFS) has been an efficient technique in exploiting multi-processors configurable characteristics. However, for large class of embedded real time applications, changing processors frequency interferes with tasks deadline guarantee. Under homogenous multiprocessor architecture, we are interested in intelligent offline/online reconfiguration based on fuzzy inference system.This paper investigates the efficiency of using fuzzy decision in multiprocessor architecture configuration. Goals of decision making are system schedulability guarantee and energy consumption optimization. Experiments concern real time periodic tasks scheduled with global EDF and realized over a developed simulation environment that supports DVFS technique. Different simulations were realized on H264 Video Decoder in order to evaluate the proposed approach response.

Energy and performance improvement using real-time DVFS for graph traversal on GPU

Graphical-processing unit (GPU) handles massive data, effectively, as compared to CPU but has high power consumption which is a primary concern recently. A real-time dynamic voltage frequency scaling technique (DVFS) to save energy as well as improve performance, and to set the core and the memory frequency at a run time are proposed. Real-time DVFS is well investigated for saving energy and improving the performance by executing a breadth-first search (BFS) graph traversal algorithm from scalable heterogeneous computing (SHOC) benchmark suite. Two modes: energy saving and performance mode, can be realised by implementation of the proposed real-time DVFS algorithm. In energy saving mode, 28.2% energy can be saved with a performance improvement of 1.4%. On the other hand, 14.2% performance improvement is accomplished while maintaining the energy saving of 11.2% in performance mode. The proposed work is realised on K40c GPU received as a grant from NVidia.

Energy and performance improvement using real-time DVFS for graph traversal on GPU

Graphical-processing unit (GPU) handles massive data, effectively, as compared to CPU but has high power consumption which is a primary concern recently. A real-time dynamic voltage frequency scaling technique (DVFS) to save energy as well as improve performance, and to set the core and the memory frequency at a run time are proposed. Real-time DVFS is well investigated for saving energy and improving the performance by executing a breadth-first search (BFS) graph traversal algorithm from scalable heterogeneous computing (SHOC) benchmark suite. Two modes: energy saving and performance mode, can be realised by implementation of the proposed real-time DVFS algorithm. In energy saving mode, 28.2% energy can be saved with a performance improvement of 1.4%. On the other hand, 14.2% performance improvement is accomplished while maintaining the energy saving of 11.2% in performance mode. The proposed work is realised on K40c GPU received as a grant from NVidia.

Power Efficient Clustering Scheme for 5G Mobile Edge Computing Environment

Mobile edge computing (MEC), which is an evolution of cloud computing, is acknowledged as a promising technology for meeting low latency and bandwidth efficiency required in fifth generation (5G) era. Accordingly, the enlargement of distributed MEC installments will be realized and their power consumption might be a significant problem in terms of operating costs for service providers. Thus, this paper proposes a theoretical framework for MEC server clustering to minimize power consumption of the MEC environment. To do this, considering power consumption behavior of MEC servers using CPUs with dynamic voltage frequency scaling, we propose a power-efficient clustering scheme (PECS) that minimizes power consumption of MEC servers by obtaining the optimal number of clusters through convex optimization. Numerical results reveal the proposed PECS reduces power consumption of the MEC environment by 12.32% relative to an existing scheme while sustaining average delay of inflows processed in MEC servers at the acceptable level without turning off MEC servers.

Dual-Loop Two-Step ZQ Calibration for Dynamic Voltage–Frequency Scaling in LPDDR4 SDRAM

This paper presents a dual-loop two-step ZQ calibration scheme with a 20-nm DRAM process to support dedicated supply voltages ( $V_{DD}$ and $V_{DDQ}$ ). The proposed calibration scheme improves system signal integrity by maintaining the targeted output level ( $V_{\mathrm {OH}}$ ) in dynamic voltage–frequency scaling (DVFS) function, where $V_{DD}$ and $V_{DDQ}$ can be independently controlled to save system power. A two-step conversion algorithm alleviates the increase in calibration time, which is caused by an additional on-die termination (ODT) calibration for command/address (CA). The offset of a dynamic comparator in a ZQ calibration engine is averaged by a fraction-referred input switching-then-averaging (FISA) scheme which minimizes the effect of kickback noise. A code-referred periodic ZQ update (CPZU) scheme tracks the voltage and temperature variation during the background mode while minimizing the chance of malfunction by interference of both DRAM internal noise and external power noise. By applying the DVFS function based on the device usage by the user scenario in this design ( $V_{DD} =1.1$ V and $V_{DDQ} = 0.95$ V) at 1.6 Gbps, the read and the write operating power consumption can be saved by 17.7% and 18.2%, respectively, without a dynamic drop in performance. Therefore, the proposed dual-loop two-step ZQ calibration design improves the overall DRAM power efficiency based on the existing LPDDR4 SDRAM architecture without circuit overhead.

Profit: <underline>Pr</underline>iority and P<underline>o</underline>wer/Per<underline>f</underline>ormance Opt<underline>i</underline>miza<underline>t</underline>ion for Many-Core Systems

As power density emerges as the main constraint for many-core systems, controlling power consumption under the thermal design power while maximizing the performance becomes increasingly critical. To dynamically save power, dynamic voltage frequency scaling techniques have proved to be effective and are widely available commercially. Meanwhile, systems have certain performance constraints that the applications should satisfy to ensure quality of service. In this paper, we present an online distributed reinforcement learning (OD-RL)-based DVFS control algorithm for many-core system performance improvement under both power and performance constraints. At the finer grain, a per-core RL method is used to learn the optimal control policy of the voltage/frequency (VF) levels in a model-free manner. At the coarser grain, an efficient global power budget reallocation algorithm is used to maximize the overall performance. The experiments show that compared to the state-of-the-art algorithms: 1) OD-RL produces up to 98% less budget overshoot; 2) up to 23% higher energy efficiency; and 3) two orders of magnitude speedup over state-of-the-art techniques for systems with hundreds of cores. Furthermore, priority-aware OD-RL can better satisfy performance constraints than OD-RL with: 1) $17.8\boldsymbol {\times }$ more epochs satisfying the performance constraints; 2) $5.6\boldsymbol {\times }$ better performance gain; and 3) $20.0\boldsymbol {\times }$ better performance-power tradeoffs under similar efficiency and scalability.

Multi‐objective approach of energy efficient workflow scheduling in cloud environments

SummaryScheduling the tasks of a workflow to the cloud resources is a well‐known N‐P hard problem. The stakeholders involved in a cloud environment have different interests in scheduling problem. In addition to the traditional objectives like makespan, budget, and deadline, optimized in workflow scheduling, considering the green aspect of cloud, (ie, energy consumption) increase the problem complexity. Moreover, the interests of a cloud's stakeholders are conflicting, and satisfying all these interests simultaneously is a big problem. In this paper, we proposed a new Multi‐Objective Genetic Algorithm(MOGA) for workflow scheduling in a cloud environment. MOGA considered the conflicting interest of the cloud stakeholders for optimization and provided a solution, which not only minimizes the makespan under the budget and deadline constraints but also provided an energy efficient solution using the dynamic voltage frequency scaling. We provided a gap search algorithm in this paper, which is used to optimize the resource utilization of the cloud's resources. We compared our results with genetic algorithms considering the budget, deadline, and energy efficiency individually. We also compared the performance of MOGA with Multi‐objective Particle Swarm Optimization (MOPSO) with the same objectives as those of MOGA. To the best of our knowledge, there is no solution presented in the literature that considers the diverse objectives considered in this work. The results show that our proposed algorithm MOGA has significantly improved not only in terms of budget, deadline, and energy but also improved the utilization of cloud's resources as compared to the competitive algorithms of this work.

Energy‐efficient task scheduling on heterogeneous computing systems by linear programming

SummaryThe continually increasing energy consumption represents a critical issue in modern heterogeneous computing systems. With the aid of dynamic voltage frequency scaling (DVFS), task scheduling is considered an effective software‐based technique for reducing the total energy consumption and minimizing the overall schedule length (makespan). A natural solution is to reclaim the slack time in a given time‐efficient schedule, which is also referred to as a “two‐pass” method or a “rescheduling” method. A number of studies have focused on slack reclamation to achieve energy reductions through heuristics; although, these methods offer suboptimal solutions. In this article, the rescheduling optimization problem is formulated as a linear program for minimizing an energy objective function subject to precedence and deadline constraints implied in the given schedule. Two types of decision variables, ie, frequency duty factors and task intervals, are defined to set up the linear model. Consequently, an optimal solution to the problem can be provided in a straightforward manner by a linear programming solver, which suggests that such a rescheduling problem belongs to the P (polynomial time) class. The experimental results show the effectiveness of the proposed approach and demonstrate that the performance is superior to that of other competitive algorithms in terms of both energy saving and runtime efficiency.

A Sub-50 µm2, Voltage-Scalable, Digital-Standard-Cell-Compatible Thermal Sensor Frontend for On-Chip Thermal Monitoring

This paper presents an on-chip temperature sensor circuit for dynamic thermal management in VLSI systems. The sensor directly senses the threshold voltage that contains temperature information using a single PMOS device. This simple structure enables the sensor to achieve an ultra-compact footprint. The sensor also exhibits high accuracy and voltage-scalability down to 0.4 V, allowing the sensor to be used in dynamic voltage frequency scaling systems without requiring extra power distribution or regulation. The compact footprint and voltage scalability enables our proposed sensor to be implemented in a digital standard-cell format, allowing aggressive sensor placement very close to target hotspots in digital blocks. The proposed sensor frontend prototyped in a 65 nm CMOS technology has a footprint of 30.1 µm2, 3σ-error of ±1.1 °C across 0 to 100 °C after one temperature point calibration, marking a significant improvement over existing sensors designed for dynamic thermal management in VLSI systems.