Energy Efficiency Of Data Centers Research Articles

Increasing the energy efficiency of a data center based on machine learning

AbstractEnergy efficiency of data centers (DCs) is of great concern due to their large amount of energy consumption and the foreseeable growth in the demand of digital services in the future. The past decade witnessed improvements of the energy efficiency of DCs from an extensive margin—a shift from small to large, more efficient DCs. Improvements from the intensive margin, that is, from more efficient operation, would be critical in limiting the energy consumption and environmental impact of DCs in the upcoming period. Machine learning algorithms have advantages in optimizing the operation of DCs to improve energy efficiency as they have shown the potential of discovering control strategies not found by traditional method, and producing working condition‐dependent control strategies. This study proposes ready‐to‐use machine learning methods with practical details to decrease the most commonly used energy efficiency metric—power usage effectiveness in DCs. We achieved an accurate prediction by properly selecting the features used in the proposed prediction models established by neural network, light gradient boosting machine, recurrent neural network, and random forests. The proposed approaches are implemented in one of the largest hyperscale DCs in China—Tencent Tianjin DC, to optimize the set points of controllable variables in the cooling system and to detect and adjust the unreasonable working conditions in the modular data centers. The lower bound of PUE reduction was 0.005 with the proposed approaches, leading to about 1500 MWh (0.24% of the total designed electricity consumption of this DC) of energy saved per year in this hyperscale DC. The proposed methods have the potential to be transferred to DCs of similar scale, and the framework of our work could serve as a guide for machine learning‐based optimization of environmental indicators in other complex product/service systems.

Simulation-based optimization of the control strategy of variable-frequency chilled water pump in data center: A case study in Beijing

Reducing the energy consumption of air-conditioning system is one of the important means to improve the operation energy efficiency of data center, and the energy consumption of water pump accounts for a relatively high proportion of the energy consumption of air-conditioning system. This paper studied the reduction of operation energy consumption of water pump from the perspective of optimizing its control strategy. Based on the simulation model of the case system, this paper simulated the operation energy consumption of the chilled water pump when two control strategies (the constant DP control and the constant DT control of the main water supply-return pipeline) were used under different load demand. The calculation results show that both two control strategies can reduce the energy consumption of the chilled water pump compared with non-control, and the constant DT control is better. In addition, under two control strategies, the load distribution characteristics of system have an impact on the working state point and energy consumption of the chilled water pump.

Strategies for Improving the Sustainability of Data Centers via Energy Mix, Energy Conservation, and Circular Energy

Information and communication technologies (ICT) are increasingly permeating our daily life and we ever more commit our data to the cloud. Events like the COVID-19 pandemic put an exceptional burden upon ICT. This involves increasing implementation and use of data centers, which increased energy use and environmental impact. The scope of this work is to summarize the present situation on data centers as to environmental impact and opportunities for improvement. First, we introduce the topic, presenting estimated energy use and emissions. Then, we review proposed strategies for energy efficiency and conservation in data centers. Energy uses pertain to power distribution, ICT, and non-ICT equipment (e.g., cooling). Existing and prospected strategies and initiatives in these sectors are identified. Among key elements are innovative cooling techniques, natural resources, automation, low-power electronics, and equipment with extended thermal limits. Research perspectives are identified and estimates of improvement opportunities are mentioned. Finally, we present an overview on existing metrics, regulatory framework, and bodies concerned.

Towards Energy Efficiency in Data Centers: An Industrial Experience Based on Reuse and Layout Changes

Data centers are widely recognized for demanding many energy resources. The greater the computational demand, the greater the use of resources operating together. Consequently, the greater the heat, the greater the need for cooling power, and the greater the energy consumption. In this context, this article aims to report an industrial experience of achieving energy efficiency in a data center through a new layout proposal, reuse of previously existing resources, and air conditioning. We used the primary resource to adopt a cold corridor confinement, the increase of the raised floor’s height, and a better direction of the cold airflow for the aspiration at the servers’ entrance. We reused the three legacy refrigeration machines from the old data center, and no new ones were purchased. In addition to 346 existing devices, 80 new pieces of equipment were added (between servers and network assets) as a load to be cooled. Even with the increase in the amount of equipment, the implementations contributed to energy efficiency compared to the old data center, still reducing approximately 41% of the temperature and, consequently, energy-saving.

A multi-dimensional double descending maximum padding priority algorithm for cloud data centers

With the development of big data technologies, green cloud data centers have become a key factor in academia and industry. An energy-efficient cloud data center can save costs for cloud computing users. However, the problem of virtual machine mapping has always been a core problem. In the most existing research, the energy consumption generated by cloud data centers has become an important bottleneck restricting the technology of cloud computing. This paper establishes a consumption model of cloud data center energy and a virtual machine mapping rule. Based on this, a multi-dimensional double descending maximum padding priority (MD3MP2) virtual machine mapping algorithm is proposed. The algorithm can not only solve the one-dimensional virtual machine mapping problem of homogeneous data centers, but also successfully solve the multi-dimensional virtual machine mapping problem of homogeneous data centers. Finally, the algorithm is compared with four other algorithms. The experimental results show that the MD3MP2 algorithm is better than the compared algorithms.

Energy and SLA-driven MapReduce Job Scheduling Framework for Cloud-based Cyber-Physical Systems

Energy consumption minimization of cloud data centers (DCs) has attracted much attention from the research community in the recent years; particularly due to the increasing dependence of emerging Cyber-Physical Systems on them. An effective way to improve the energy efficiency of DCs is by using efficient job scheduling strategies. However, the most challenging issue in selection of efficient job scheduling strategy is to ensure service-level agreement (SLA) bindings of the scheduled tasks. Hence, an energy-aware and SLA-driven job scheduling framework based on MapReduce is presented in this article. The primary aim of the proposed framework is to explore task-to-slot/container mapping problem as a special case of energy-aware scheduling in deadline-constrained scenario. Thus, this problem can be viewed as a complex multi-objective problem comprised of different constraints. To address this problem efficiently, it is segregated into three major subproblems (SPs), namely, deadline segregation, map and reduce phase energy-aware scheduling. These SPs are individually formulated using Integer Linear Programming. To solve these SPs effectively, heuristics based on Greedy strategy along with classical Hungarian algorithm for serial and serial-parallel systems are used. Moreover, the proposed scheme also explores the potential of splitting Map/Reduce phase(s) into multiple stages to achieve higher energy reductions. This is achieved by leveraging the concepts of classical Greedy approach and priority queues. The proposed scheme has been validated using real-time data traces acquired from OpenCloud. Moreover, the performance of the proposed scheme is compared with the existing schemes using different evaluation metrics, namely, number of stages, total energy consumption, total makespan, and SLA violated. The results obtained prove the efficacy of the proposed scheme in comparison to the other schemes under different workload scenarios.

A comparative study of energy and task efficient load balancing algorithms in cloud computing

The future growth of the Internet of Services has fundamentally changed the emergence of cloud computing. Cloud data centres serve multiple tenant demands for cloud applications that discharge vast amounts of electricity, leading to high operating costs and environmental diffusion of carbon dioxide (CO2). To fix this is the need for preservation to enable potential use by building a new structure and measuring the effect in a cloud data centre. Consequently, the use of pruned electricity reduces the cost of processing power. In order to meet energy-efficient data centres in the cloud, adjusting to optimal load balancing processing a good way for energy savings. To minimise the large energy use of cloud data centres herds, this focuses on increasing efficacy by breaking the workload evenly. In this paper, we plan to provide a comprehensive comparative analysis in cloud computing of current load balancing algorithms.

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.

QoS Promotion in Energy-Efficient Datacenters Through Peak Load Scheduling

To build energy-efficient datacenters, one widely used way is to dynamically manipulate the quantity of available hardware resources on demand. However, when bursty workloads appear, additional time overhead is required for resource gearing, thus incurring a performance degradation. To explore this problem, we present an intelligible analysis on the request handing in a VM. From the analysis, we find that, when workloads are overloaded (i.e., peak loads appear), the number of QoS guaranteed requests can be greatly increased by deferring the scheduling of a few requests. Inspired by this finding, we propose a Peak Load Scheduling Control (PLSC) method to promote the Quality of Service (QoS) of peak loads for modern energy-efficient datacenters. However, peak loads are usually difficult to identify. To overcome this difficulty, PLSC tracks the number of requests residing in a VM by leveraging a two-tier request queue maintained by it. When the number exceeds the capability of the VM, it means that peak loads appear. In this case, PLSC adds some delay-tolerant requests to the secondary queue. The scheduling of requests in the secondary queue is controlled with a lower priority than that of requests in the primary queue. Sequentially, with critical requests maintained in the primary queue, PLSC shortens the response time of critical requests. In addition, PLSC expands the number of QoS-guaranteed requests. Comprehensive experiments are conducted to attest the effectiveness of PLSC, by simulating a typical energy-efficient datacenter. The experimental results show that PLSC significantly promotes the QoS of workloads with a negligible impact on energy saving.

Energy Efficient Placement of Workloads in Composable Data Center Networks

This paper studies the energy efficiency of composable data center (DC) infrastructures over network topologies. Using a mixed integer linear programming (MILP) model, we compare the performance of disaggregation at rack-scale and pod-scale over selected electrical, optical and hybrid network topologies relative to a traditional DC. Relative to a pod-scale DC, the results show that physical disaggregation at rack-scale is sufficient for optimal efficiency when the optical network topology is adopted, and resource components are allocated in a suitable manner. The optical network topology also enables optimal energy efficiency in composable DCs. The paper also studies logical disaggregation of traditional DC servers over an optical network topology. Relative to physical disaggregation at rack-scale, logical disaggregation of server resources within each rack enables marginal fall in the total DC power consumption (TDPC) due to improved resource demands placement. Hence, an adaptable composable infrastructure that can support both in memory (access) latency sensitive and insensitive workloads is enabled. We also conduct a study of the adoption of micro-service architecture in both traditional and composable DCs. Our results show that increasing the modularity of workloads improves the energy efficiency in traditional DCs, but disproportionate utilization of DC resources persists. A combination of disaggregation and micro-services achieved up to 23% reduction in the TDPC of the traditional DC by enabling optimal resources utilization and energy efficiencies. Finally, we propose a heuristic for energy efficient placement of workloads in composable DCs which replicates the trends produced by the MILP model formulated in this paper.

An experimental and numerical investigation of novel solution for energy management enhancement in data centers using underfloor plenum porous obstructions

This investigation focuses on improving the airflow distribution in data centers. In many data centers vortices form in the plenum during operations. These vortices cause spatial and temporal non-uniformities and may give rise to hot regions in the data center which in turn impacts performance and reliability of the IT equipment. The current study identifies a novel approach using porous partitions in the plenum and demonstrates a significant generalized approach that is easily adoptable in existing and future data centers. For improving the overall data center energy efficiency and the cooling system effectiveness by eliminating a critical source of inefficiency. The results of quantitative and qualitative analyses of the underfloor plenum pressure field, perforated tiles airflow rate, and air temperature at the rack intake side with and without partitions are presented. Different data center configurations are studied using physics-based experimentally validated Computational Fluid Dynamics (CFD) model. The CFD model results showed that the partitions eliminated the presence of vortices in the underfloor plenum and thus enabled a more uniform pressure distribution and tile airflow delivery. Regarding rack inlet temperature, the results showed that the partitions significantly improved the air temperature at the rack inlet. Finally, a geometrical parametric study is performed. An ideal design is and demonstrated to improve the Supply Heat Index (SHI) by about 10%, while the amount of IT equipment that exceeded the ASHRAE recommended supply air temperature (SAT) was reduced by about 40%, and the floor leakage was cut in half.

Energy efficient virtual machine migration approach with SLA conservation in cloud computing

In the age of online workload explosion, cloud users are increasing exponentialy. Therefore, large scale data centers are required in cloud environment that leads to high energy consumption. Hence, optimal resource utilization is essential to improve energy efficiency of cloud data center. Although, most of the existing literature focuses on virtual machine (VM) consolidation for increasing energy efficiency at the cost of service level agreement degradation. In order to improve the existing approaches, load aware three-gear THReshold (LATHR) as well as modified best fit decreasing (MBFD) algorithm is proposed for minimizing total energy consumption while improving the quality of service in terms of SLA. It offers promising results under dynamic workload and variable number of VMs (1–290) allocated on individual host. The outcomes of the proposed work are measured in terms of SLA, energy consumption, instruction energy ratio (IER) and the number of migrations against the varied numbers of VMs. From experimental results it has been concluded that the proposed technique reduced the SLA violations (55%, 26% and 39%) and energy consumption (17%, 12% and 6%) as compared to median absolute deviation (MAD), inter quartile range (IQR) and double threshold (THR) overload detection policies, respectively.

Energy, exergy, and economic analysis of a data center energy system driven by the CO2 ground source heat pump: Prosumer perspective

With increasing reliance on information technology by modern society, the scale and energy consumption of data centers (DCs) are rising rapidly. Approximately 40% of the energy consumed by DCs is associated with refrigeration systems used to maintain optimal operation temperatures. Such systems produce large amounts of waste heat; therefore, the utilization of the waste heat is essential for the optimization of the energy efficiency of DCs, especially with regard to DC prosumers (producers and consumers). However, the low temperature of the waste heat and variation in load demand are two factors limiting the exploitation of waste heat from DCs. In this study, a prosumer DC energy system based on a ground source heat pump (GSHP) is proposed for the recycling of waste heat from DCs for the heating of surrounding buildings. The heat generated in the non-heating period is stored in the soil and made available for building heating using a carbon dioxide (CO2) direct-expansion GSHP. Furthermore, the cooling power consumption, total power consumption, exergy efficiency, and net profit of the prosumer DC waste heat energy recovery system are evaluated. Compared to conventional cooling systems that use air source heat pumps (ASHPs) for building heating, the energy, exergy, and economic (3E) performance of the prosumer DC have the following advantages: 1) the prosumer DC waste heat energy recovery system yields a heat-power ratio of 5.70, with a total power consumption approximately half (52.86%) that of the conventional system; 2) the matching thermodynamic properties are greatly improved; specifically, the exergy efficiency of the prosumer DC waste heat energy recovery system is 3.87 fold that of the conventional system; 3) better economic revenue of the prosumer DC waste heat energy recovery system generates more economic revenue, with an annual net profit approximately 190.34% of that generated by the conventional system. In general, the prosumer system facilitates flexible and effective utilization of the low-temperature waste heat, by coupling the upstream power supplier and downstream thermal users through a GSHP.

Combined particle swarm optimization and Ant Colony System for energy efficient cloud data centers

AbstractOne of the foremost issues deliberated by the cloud infrastructure providers is to minimize the costs by reducing energy consumption. Though many heuristic algorithms such as Genetic Algorithm (GA), Simulated Annealing (SA), and so on, have been existed for task scheduling or server consolidation to reduce energy consumption, more energy savings in data centers is required. In this paper, a combination of two distinctive heuristic algorithms, namely Particle Swarm Optimization (PSO) technique for proficient task scheduling and Ant Colony System (ACS) for effective server consolidation are proposed. In PSO, each particle in the swarm finds a task scheduling solution and updates the best solution. It is faster than many meta‐heuristic algorithms. In ACS, each ant in the ant colony tries to find the best destination PMs for migrating VMs from overloaded or underloaded PMs. The proposed PSO‐ACS algorithm is experimented with PlanetLab real workload traces available in the CloudSim toolkit. The algorithm provides reduction in energy consumption, number of VM migrations, and Energy SLA Violation (ESV) metrics while maintaining QoS. In the best‐case scenario, the proposed method reduces the ESV metric by 28%, 27%, and 18% in comparison with other algorithms like LR, ACS, and PSO‐LR.

Genetic algorithm based cooling energy optimization of data centers

PurposeThis study aims to present development of genetic algorithm (GA)-based framework aimed at minimizing data center cooling energy consumption by optimizing the cooling set-points while ensuring that thermal management criteria are satisfied.Design/methodology/approachThree key components of the developed framework include an artificial neural network-based model for rapid temperature prediction (Athavale et al., 2018a, 2019), a thermodynamic model for cooling energy estimation and GA-based optimization process. The static optimization framework informs the IT load distribution and cooling set-points in the data center room to simultaneously minimize cooling power consumption while maximizing IT load. The dynamic framework aims to minimize cooling power consumption in the data center during operation by determining most energy-efficient set-points for the cooling infrastructure while preventing temperature overshoots.FindingsResults from static optimization framework indicate that among the three levels (room, rack and row) of IT load distribution granularity, Rack-level distribution consumes the least cooling power. A test case of 7.5 h implementing dynamic optimization demonstrated a reduction in cooling energy consumption between 21%–50% depending on current operation of data center.Research limitations/implicationsThe temperature prediction model used being data-driven, is specific to the lab configuration considered in this study and cannot be directly applied to other scenarios. However, the overall framework can be generalized.Practical implicationsThe developed framework can be implemented in data centers to optimize operation of cooling infrastructure and reduce energy consumption.Originality/valueThis paper presents a holistic framework for improving energy efficiency of data centers which is of critical value given the high (and increasing) energy consumption by these facilities.

Toward Enhancing the Energy Efficiency and Minimizing the SLA Violations in Cloud Data Centers

Recently, the problem of Virtual Machine Placement (VMP) has received enormous attention from the research community due to its direct effect on the energy efficiency, resource utilization, and performance of the cloud data center. VMP is considered as a multidimensional bin packing problem, which is a type of NP-hard problem. The challenge in VMP is how to optimally place multiple independent virtual machines into a few physical servers to maximize a cloud provider’s revenue while meeting the Service Level Agreements (SLAs). In this paper, an effective multiobjective algorithm based on Particle Swarm Optimization (PSO) technique for the VMP problem, referred to as VMPMOPSO, is proposed. The proposed VMPMOPSO utilizes the crowding entropy method to optimize the VMP and to improve the diversity among the obtained solutions as well as accelerate the convergence speed toward the optimal solution. VMPMOPSO was compared with a simple single-objective algorithm, called First-Fit-Decreasing (FFD), and two multiobjective ant colony and genetic algorithms. Two simulation experiments were conducted to verify the effectiveness and efficiency of the proposed VMPMOPSO. The first experiment shows that the proposed algorithm has better performance than the algorithms we compared it to in terms of power consumption, SLA violation, and resource wastage. The second indicates that the Pareto optimal solutions obtained by applying VMPMOPSO have a good distribution and a better convergence than the comparative algorithms.

BiTE: a dynamic bi-level traffic engineering model for load balancing and energy efficiency in data center networks

With the recent significant growth of virtualization and cloud services, the data center network (DCN) as the underlying infrastructure is more important. The increasing and changing volume of workloads highlights critical issues such as load balancing and energy efficiency in data centers. Large path diversity in DCNs introduces multipath forwarding as a promising approach to improve load distribution. However, the over-provisioned DCNs consume large amounts of power while the network is under full capacity most of the time. Accordingly, this paper proposes BiTE, a dynamic bi-level traffic engineering (TE) scheme in a hierarchical Software Defined Networking (SDN)-based DCN to strike a balance between load balancing and energy efficiency objectives. BiTE consists of decision-making problems at two levels modeled as a multi-period bi-level optimization problem, where each decision maker optimizes one of objectives. According to the inherent complexity of bi-level programming, a co-evolutionary metaheuristic algorithm is proposed for solving BiTE. BiTE performance is evaluated in comparison to NSGA-II algorithm and four previously proposed TE schemes in terms of several load balancing and energy saving metrics under different scenarios. The results show that BiTE performs well in traffic load balancing while preserves the energy efficiency. We apply the Analytic Hierarchy Process (AHP) method to multi-criteria analyze and rank the performance of studied TE mechanisms. AHP results for different scenarios indicate that BiTE is in first or second place in terms of the overall performance score among six studied approaches.

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.

A Novel Machine Learning-Based Load-Adaptive Power Supply System for Improved Energy Efficiency in Datacenters

A Novel Machine Learning-Based Load-Adaptive Power Supply System for Improved Energy Efficiency in Datacenters

Design and Development of an Energy Efficient Multimedia Cloud Data Center with Minimal SLA Violation

Design and Development of an Energy Efficient Multimedia Cloud Data Center with Minimal SLA Violation