Improve Data Center Energy Efficiency Research Articles

Energy savings in direct air-side free cooling data centers: A cross-system modeling and optimization framework

The advancement of intelligent technology has led to a significant increase in the electricity consumption of data centers. Efficient free cooling and cross-system optimization are major means to improve data center energy efficiency. This work aims to develop a cross-system modeling and optimization methodology to reduce the energy consumption within data centers equipped with direct air-side economizer. A series of physics-based models are built, accounting for free cooling costs, supply air conditions (including both temperature and humidity), and outdoor weather fluctuations. A holistic energy optimization framework is proposed to determine optimal system setpoints that accommodate changes in computing workload arrival rates and outdoor environments. Experimental results demonstrate that the optimal cost characteristic curve for free cooling can be represented as a piecewise function. The advantages of the proposed method can save up to more than 5% of overall energy, which occurs during summer weeks, and more than 46% of cooling energy during winter weeks. The highest share of cooling energy consumption is the supply air fan, reaching up to more than 92% in some situations. The operation of a direct air-side economizer is found to contain four optimal modes.

Optimizing Energy Efficiency for Data Center via Parameterized Deep Reinforcement Learning

The rapid advancements in cloud computing, Big Data and their related applications have led to a skyrocketing increase in data center energy consumption year by year. The prior approaches for improving data center energy efficiency mostly suffer from high system dynamics or the complexity of data centers. In this paper, we propose an optimization framework based on deep reinforcement learning, named DeepEE, to jointly optimize energy consumption from the perspectives of task scheduling and cooling control. In DeepEE, a PArameterized action space based Deep Q-Network (PADQN) algorithm is proposed to tackle the hybrid action space problem. Then, a dynamic time factor mechanism for adjusting cooling control interval is introduced into PADQN (PADQN-D) to achieve more accurate and efficient coordination of IT and cooling subsystems. Finally, in order to train and evaluate the proposed algorithms safely and quickly, a simulation platform is built to model the dynamics of IT and cooling subsystems. Extensive real-trace based experiments illustrate that: 1) the proposed PADQN algorithm can save up to 15% and 10% energy consumption compared with the baseline siloed and joint optimization approaches respectively; 2) the proposed PADQN-D algorithm with dynamic cooling control interval can better adapt to the change of IT workload; 3) our proposed algorithms achieve more stable performance gain in terms of power consumption by adopting the parameterized action space.

Cooperatively Improving Data Center Energy Efficiency Based on Multi-Agent Deep Reinforcement Learning

The problem of high power consumption in data centers is becoming more and more prominent. In order to improve the energy efficiency of data centers, cooperatively optimizing the energy of IT systems and cooling systems has become an effective way. In this paper, a model-free deep reinforcement learning (DRL)-based joint optimization method MAD3C is developed to overcome the high-dimensional state and action space problems of the data center energy optimization. A hybrid AC-DDPG cooperative multi-agent framework is devised for the improvement of the cooperation between the IT and cooling systems for further energy efficiency improvement. In the framework, a scheduling baseline comparison method is presented to enhance the stability of the framework. Meanwhile, an adaptive score is designed for the architecture in consideration of multi-dimensional resources and resource utilization improvement. Experiments show that our proposed approach can effectively reduce energy for data centers through the cooperative optimization while guaranteeing training stability and improving resource utilization.

A multi-objective Monarch Butterfly Algorithm for virtual machine placement in cloud computing

The growing demand for cloud computing adoption presents more challenges for researchers to make cloud computing more efficient and affordable for infrastructure providers and end users. The management of cloud computing involves investing in IT infrastructure in the first phase and investing in energy, maintenance and space costs in the second phase. However, energy costs account for a large portion of cloud management costs, and saving energy consumption can significantly reduce overall cloud management costs. Server consolidation is a strategy to improve data center energy efficiency and resource utilization. Virtual machine (VM) placement is considered one of the main problems with VM consolidation. The VM placement problem aims to reduce the number of active physical machines in data center to reduce data center power consumption and maintenance costs. However, the waste of data center resources has a significant impact on the energy efficiency of the data center, so it should be considered in the VM placement strategy. This paper proposes a new method based on the Monarch Butterfly Optimization algorithm (MBO) called MBO-VM for new virtual machine placement, designed to maximize packaging efficiency and reduce the number of active physical servers. CloudSim toolkit is used to test the efficiency of the proposed MBO-VM approach under real cloud workloads as well as synthetic workloads. Simulation results show that MBO-VM produces significantly better results compared with known VM placement techniques. The proposed MBO-VM can reduce the number of active servers more effectively and maximize the packaging efficiency.

CPicker: Leveraging Performance-Equivalent Configurations to Improve Data Center Energy Efficiency

The poor energy proportionality of server is seen as the principal source for low energy efficiency of modern data centers. We find that different resource configurations of an application lead to similar performance, but have distinct energy consumption. We call this phenomenon as “performance-equivalent resource configurations (PERC)”, and its performance range is called equivalent region (ER). Based on PERC, one basic idea for improving energy efficiency is to select the most efficient configuration from PERC for each application. However, it cannot support every application to obtain optimal solution when thousands of applications are run simultaneously on resource-bounded servers. Here we propose a heuristic scheme, CPicker, based on genetic programming to improve energy efficiency of servers. To speed up convergence, CPicker initializes a high quality population by first choosing configurations from regions that have high energy variation. Experiments show that CPicker obtains above 17% energy efficiency improvement compared with the greedy approach, and less than 4% efficiency loss compared with the oracle case.

Improving Data Center Energy Efficiency With Advanced Thermal Management

Experimental investigation of data center cooling and computational energy efficiency improvement through advanced thermal management was performed. A chiller-less data center liquid cooling system was developed that transfers the heat generated from computer systems to the outdoor ambient environment while eliminating the need for energy-intensive vapor-compression refrigeration. This liquid cooling system utilizes a direct-attach cold-plate approach that enables the use of warm water at temperature a few degrees above outdoor ambient to achieve lower chip junction temperatures than refrigerated air. Using this approach, we demonstrated a cooling energy reduction by over 90% and computational energy reduction of up to 14% compared to traditional refrigerated air cooled data centers. To enable future computational efficiency improvements through high-density 3-D-chip stacking, we developed a 3-D compatible chip-embedded two-phase liquid cooling technology where a dielectric coolant is pumped through microscale cavities to provide thermal management of chips within the stack. In two-phase cooling, liquid is converted to vapor, which increases the capacity to remove heat, while the dielectric fluid enables integration with chip electrical interconnects. A test vehicle simulating an eight-core microprocessor was fabricated with embedded cooling channels. Results demonstrate that this volumetrically efficient cooling solution compatible with 3-D chip stacks can manage three times the core power density of today's high-power processor while maintaining the device temperature well within limits.

Improving Data Center Energy Efficiency Using a Cyber-physical Systems Approach: Integration of Building Information Modeling and Wireless Sensor Networks

The increase in data center operating costs is driving innovation to improve their energy efficiency. Previous research has investigated computational and physical control intervention strategies to alleviate the competition between energy consumption and thermal performance in data center operation. This study contributes to the body of knowledge by proposing a cyber-physical systems (CPS) approach to innovatively integrate building information modeling (BIM) and wireless sensor networks (WSN). In the proposed framework, wireless sensors are deployed strategically to monitor thermal performance parameters in response to runtime server load distribution. Sensor data are collected and contextualized in reference to the building information model that captures the geometric and functional characteristics of the data center, which will be used as inputs of continuous simulations aiming to predict real-time thermal performance of server working environment. Comparing the simulation results against historical performance data via machine learning and data mining, facility managers can quickly pinpoint thermal hot zones and actuate intervention procedures to improve energy efficiency. This BIM-WSN integration also facilitates smarter power management by capping runtime power demand within peak power capacity of data centers and alerting power outage emergencies. This paper lays out the BIM-WSN integration framework, explains the working mechanism, and discusses the feasibility of implementation in future work.

Long-term transient thermal analysis using compact models for data center applications

Reduced-order thermal models are necessary to enable real-time assessment of the optimum operating and control conditions to improve data center energy efficiency. A 3D computational fluid dynamics (CFD) and heat transfer model of a basic raised floor, air cooled, hot aisle/cold aisle data center configuration was developed and simulation results were used to generate snapshots for initializing compact models based on the Proper Orthogonal Decomposition (POD) and the Nonlinear Principle Component Analysis (NLPCA) methods. The specific focus of the study was to numerically investigate how the thermal trends can be affected by the long-term transient flow patterns associated with leakage and how the compact models (e.g. POD and NLPCA) can be utilized for much faster implementations and characterizations of the transient flows. Using both the POD and the NLPCA method, good agreement was achieved between the full simulation results and the compact models for predicting the dynamically developed local flow structure over a range of transient cooling air supply operating conditions. In addition, the NLPCA method was implemented to better characterize the nonlinear aspects of the CFD results. The benefits of using both the POD and NLPCA methods are discussed in relation to constructing compact models as real-time predictive tools. A systematic use of the compact models has also been proposed, to enable more robust thermal management and control of data centers.

Opportunities and challenges to unify workload, power, and cooling management in data centers

Independent optimization for workload and power management, and active cooling control have been studied extensively to improve data center energy efficiency. Recently, proposals have started to advocate unified workload, power, and cooling management for further energy savings. In this paper, we study this problem with the objectives of both saving energy and capping power. We present the detailed models derived in our previous work from experiments on an blade enclosure system that can be representative of a data center, discuss the optimization opportunities for coordinated power and cooling management, and the challenges for controller design. We then propose a few design principles and examples for unified workload management, power minimization, and power capping. Our simulation-based evaluation shows that the controllers can cap the total power consumption while maintaining the thermal conditions and improve the overall energy efficiency. We argue that the same opportunities, challenges, and designs are also generally applicable to data center level management.