Data Center Load Research Articles

Numerical simulation of a hybrid energy system proposed for low carbon data center

Low-carbon energy application for both large grid and distributed energy consumers is an urgent problem to be solved. Data centers (DC) are the typical distributed large-size energy consumers, and the application of renewable energy and energy storage is a promising solution for data center decarbonization. This paper developed a complete hybrid energy system (HES) simulation process for data center applications. The photovoltaic (PV) cell, hydrogen electrolysis, hydrogen storage, proton exchange membrane fuel cell (PEMFC), lithium battery (BAT), grid, and data center load are considered in the HES. A one-dimensional two-phase flow PEMFC model is developed. The detailed mass transfer process inside the PEMFC cell is modeled. The PEMFC model-solving algorithm is developed and integrated into the HES simulation program. The energy management system (EMS) based on a logical flow chart energy management strategy is integrated by editable subfunction. The HES simulation program is coded and implemented based on the MATLAB-SIMULINK environment. The yearly HES performance simulations are conducted and analyzed for a 200 kW data center in Xi’an with different HES parameters and an appropriate HES parameter arrangement is provided. The main contribution of this work is the development of an editable HES simulation program that is extensible for data center applications.

Task Offloading in Real-Time Distributed Energy Power Systems

The distributed energy power system needs to provide sufficient and flexible computing power on demand to meet the increasing digitization and intelligence requirements of the smart grid. However, the current distribution of the computing power and loads in the energy system is unbalanced, with data center loads continuously increasing, while there is a large amount of idle computing power at the edge. Meanwhile, there are a large number of real-time computing tasks in the distributed energy power system, which have strict requirements on execution deadlines and require reasonable scheduling of multi-level heterogeneous computing power to meet real-time computing demands. Based on the aforementioned background and issues, this paper studies the real-time service scheduling problem in a multi-level heterogeneous computing network of distributed energy power systems. Specifically, we consider the divisibility of tasks in the model. This paper presents a hierarchical real-time task-scheduling framework specifically designed for distributed energy power systems. The framework utilizes an orchestrating agent (OA) as the execution environment for the scheduling module. Building on this, we propose a hierarchical selection algorithm for choosing the appropriate network layer for real-time tasks. Further, we develop two scheduling algorithms based on greedy strategy and genetic algorithm, respectively, to effectively schedule tasks. Experiments show that the proposed algorithms have a superior success rate in scheduling compared to other current algorithms.

Study of dynamic performance of PEMFC-based CCHP system in a data center based on real-time load and a novel synergistic control method with variable working conditions

In this paper, the dynamic performance of Proton Exchange Membrane Fuel Cell-based Combined Cooling Heating and Power (PEMFC-based CCHP) system combined with a data center real-time load is studied. A novel synergistic control method is proposed to achieve optimization of energy flow between system units under fluctuating load. First, the off-design working condition models of PEMFC-based CCHP system are constructed. The delay in the changes of data center temperature is considered because of the thermal inertia under dynamic conditions. Second, the dynamic performance of PEMFC-based CCHP system to step changes of data center load is analyzed. Third, to further reveal the energy flow between units, the units output under varying load conditions are compared under the constraints of power and refrigeration supply. Finally, the dynamic performance of PEMFC-based CCHP system under the traditional control method of following electric load are revealed. A novel synergistic control method for the PEMFC-based CCHP system is proposed by refining the traditional method to optimize the energy flow between units. The result presents that the average energy efficiency reaches 78.29 %. Compared with the following electric load method, the energy efficiency is increased by 6.02 %. The response time of synergistic control method is 27.0 ms. Improved energy efficiency and fast response have been achieved.

Алгоритм динамического распределения и балансировки нагрузки в распределенных облачных вычислениях

A mathematical model and algorithm of a two-level load management system for virtual clusters of a data processing center (data center) have been developed. At the first management level, virtual machines (VMs) are assigned to physical servers. At the same time, a greedy algorithm is used with restrictions on the time of searching for acceptable load distribution alternatives. The second level of management is implemented taking into account the chaotic structure of network traffic between the data center and users. Checking for the randomness of a time series of information traffic is carried out using Lyapunov exponents. The predictive model of the load intensity is implemented using the method of phase space reconstruction based on a set of values of a one-dimensional time series. When constructing a reconstructed phase space attractor, the time delay value is selected from the condition of reaching the zero value of the autocorrelation function, and the dimension of the embedding is determined by the angle of inclination of the straight line approximating the dependence of the value of the correlation integral on the radius of a given threshold point. The Tayler window is used to exclude correlated points in the numerical series. The criterion for evaluating the effectiveness of the developed algorithm is an integral indicator of the deviation of the load of each server from a given level. The proposed model can be used to build a data center load balancing system in conditions of its nonlinear nature.

Data Center Energy Evaluation Tool Development and Analysis of Power Usage Effectiveness with Different Economizer Types in Various Climate Zones

Data centers are energy-intensive facilities, with over 95% of their total cooling load attributed to the heat generated by information technology equipment (ITE). Various energy-saving techniques have been employed to enhance data center efficiency and to reduce power usage effectiveness (PUE). Among these, economizers using outdoor air for cooling are the most effective for addressing year-round cooling demands. Despite the simplicity of the load composition, analyzing data center cooling systems involves dynamic considerations, such as weather conditions, system conditions, and economizer control. A PUE interpretation tool was specifically developed for use in data centers, aimed at addressing the simplicity of data center loads and the complexity of system analysis. The tool was verified through a comparison with results from DesignBuilder implementing the EnergyPlus algorithm. Using the developed tool, a comparative analysis of economizer strategies based on the PUE distribution was conducted, with the aim of reducing the PUE of data centers across various climatic zones. The inclusion of evaporative cooling (EC) further improved cooling efficiency, leading to reductions in PUE by approximately 0.02 to 0.05 in dry zones. Additionally, wet zones exhibited PUE reductions, ranging from approximately 0.03 to 0.07, with the implementation of indirect air-side economizer (IASE). Sensitivity and uncertainty analysis were further conducted. The computer room air handler (CRAH) supply temperature and CRAH temperature difference were the most influential factors affecting the annual PUE. For the direct air-side economizer (DASE) and DASE + EC systems, higher PUE uncertainty was observed in zones 1B, 3B, 4B, and 5B, showing ranges of 1.17–1.39 and 1.15–1.17, respectively. In the case of the IASE and IASE + EC systems, higher PUE uncertainty was noted in zones 0A, 0B, 1A, 1B, and 2A, with ranges of 1.22–1.43 and 1.17–1.43, respectively. The distinctive innovation of the tool developed in this study is characterized by its integration of specific features unique to data centers. It streamlines the computation of cooling loads, thus minimizing the burden of input, and delivers energy consumption data for data center cooling systems with a level of precision comparable to that of commercial dynamic energy analysis tools. It provides data center engineers with a valuable resource to identify optimal alternatives and system design conditions for data centers. This empowers them to make informed decisions based on energy efficiency enhancements, thereby strengthening their ability to improve energy efficiency.

Spatio-temporal management of renewable energy consumption, carbon emissions, and cost in data centers

Under the background of "carbon neutrality ", data center enterprises are confronted with the challenges of high energy costs and the need to manage carbon emissions. Compared with traditional energy sources, renewable energy possesses the advantages of being low-carbon and cost-effective, making it an essential avenue for data centers to enhance their utilization of renewable energy. By employing a spatio-temporal scheduling method for computing power load, data center enterprises can maximize the benefits of renewable energy, achieve low-carbon and cost-effective operation, and enhance the consumption of renewable energy. This study developed a spatio-temporal scheduling model for computing load in data centers, with a specific focus on optimizing the utilization of renewable energy while considering the goals of low-carbon emissions and cost-effectiveness. A two-stage spatio-temporal scheduling algorithm (ESTS) was designed and implemented, and three sets of experiments were conducted to assess the effectiveness and applicability of offline load scheduling using offline load data from Alibaba's cluster-trace-v2018. The results demonstrate that the proposed scheduling method can achieve a significant reduction of carbon emissions by 70% and operating costs by 40% across various scenarios. Moreover, during the summer season when renewable energy is abundant, the application of this scheduling method in a single data center can effectively achieve the objectives of managing low-carbon emissions and minimizing costs.

Spatio-temporal load migration potential of data centers: Evaluation and application

The wide distribution of data centers and the delay tolerance of computing tasks endow data center loads with adjustable characteristics in both temporal and spatial dimensions. Due to the characteristics of abundant spatiotemporal flexibility, data centers can participate in the optimization of power system operation and regulation. To quantify this flexibility, this paper proposes a spatiotemporal load migration potential evaluation model. Meanwhile, a data center energy management strategy is combined with the characteristics of spatiotemporal load migration, which deeply exploits the migration potential through the spatiotemporal redistribution of delay-tolerant tasks, as well as server ON/OFF scheduling and CPU operating frequency scaling across different spatial locations. A case study demonstrates that adopting the proposed approach considering an energy management strategy can effectively improve the load migration potential of data centers. The migration characteristics of data centers have great application prospects in reducing carbon emissions and enhancing operational flexibility.

Optimizing Cloud Computing Applications with a Data Center Load Balancing Algorithm

Delivering scalable and on-demand computing resources to users through the usage of the cloud has become a common paradigm. The issues of effective resource utilisation and application performance optimisation, however, become more pressing as the demand for cloud services rises. In order to ensure efficient resource allocation and improve application performance, load balancing techniques are essential in dispersing incoming network traffic over several servers. The workload balancing in the context of cloud computing, particularly in the Infrastructure as a Service (IaaS) model, continues to be difficult. Due to available virtual machines and the limited resources, efficient job allocation is essential. To prevent prolonged execution delays or machine breakdowns, cloud service providers must maintain excellent performance and avoid overloading or underloading hosts. The importance of task scheduling in load balancing necessitates compliance with Service Level Agreement (SLA) standards established by cloud developers for consumers. The suggested technique takes into account Quality of Service (QoS) job parameters, VM priorities, and resource allocation in order to maximise resource utilisation and improve load balancing. The proposed load balancing method is in line with the results in the body of existing literature by resolving these problems and the current research gap. According to experimental findings, the Dynamic LBA algorithm currently in use is outperformed by an average resource utilisation of 78%. The suggested algorithm also exhibits excellent performance in terms of accelerated Makespan and decreased execution time.

Energy saving strategy of cloud data computing based on convolutional neural network and policy gradient algorithm.

Cloud Data Computing (CDC) is conducive to precise energy-saving management of user data centers based on the real-time energy consumption monitoring of Information Technology equipment. This work aims to obtain the most suitable energy-saving strategies to achieve safe, intelligent, and visualized energy management. First, the theory of Convolutional Neural Network (CNN) is discussed. Besides, an intelligent energy-saving model based on CNN is designed to ameliorate the variable energy consumption, load, and power consumption of the CDC data center. Then, the core idea of the policy gradient (PG) algorithm is introduced. In addition, a CDC task scheduling model is designed based on the PG algorithm, aiming at the uncertainty and volatility of the CDC scheduling tasks. Finally, the performance of different neural network models in the training process is analyzed from the perspective of total energy consumption and load optimization of the CDC center. At the same time, simulation is performed on the CDC task scheduling model based on the PG algorithm to analyze the task scheduling demand. The results demonstrate that the energy consumption of the CNN algorithm in the CDC energy-saving model is better than that of the Elman algorithm and the ecoCloud algorithm. Besides, the CNN algorithm reduces the number of virtual machine migrations in the CDC energy-saving model by 9.30% compared with the Elman algorithm. The Deep Deterministic Policy Gradient (DDPG) algorithm performs the best in task scheduling of the cloud data center, and the average response time of the DDPG algorithm is 141. In contrast, the Deep Q Network algorithm performs poorly. This paper proves that Deep Reinforcement Learning (DRL) and neural networks can reduce the energy consumption of CDC and improve the completion time of CDC tasks, offering a research reference for CDC resource scheduling.

Series Editorial: Cloud and Edge Computing

Cloud and Edge Computing are currently undergoing a sub-stantive transformation on several fronts, from applications to hardware, and from architectures to devices. This issue of the Cloud and Edge Computing Series solicited articles in the area of Cloud and Edge Computing to address the main issues concerned with evolving processes, supporting pedagogies, and applications in cloud computing, networking, and storages technologies. There have been advances on both the cloud and edge computing fronts that have affected this line of technology. Paradigms, such as computing in virtualization-based architectures, issues on geo-graphical constraints for deploying clouds, and the use of SDN/NFV in clouds, for example, were of special interest for this issue of the Series. Nevertheless, our attention was also focused on data center network (DCN) architectures, security, load balancing, and application data streaming supported by evidence from simulations, analysis, or experiments. Cloud and Edge Computing are also incorporated with the Internet of Things (IoT) ecosystem. IoT requiring multiple access edge computing systems is gaining momentum and considered to be deployed in smart cities, public safety, and e-health. The last, but not least important, factor we expected to see in articles was the standardization aspects of Cloud and Edge Computing technology. We were especially interested in articles that could address communications standards in networking for clouds, fogs, and edge computing.

Using geographic load shifting to reduce carbon emissions

An increasing focus on the electricity use and carbon emissions associated with computing has lead to pledges by major cloud computing companies to lower their carbon footprint. Data centers have the unique ability to shift computing load between different geographical locations, giving rise to flexibility that can be employed to reduce carbon emissions. In this paper, we present a model where data centers shift load independently of the ISOs. We first consider the impact of load shifting guided by locational marginal carbon emissions, λ CO 2 , a sensitivity metric that measures the impact of incremental load shifts. Relative to previous models for data center load shifting, the presented model improves accuracy and includes more realistic assumptions regarding the operation of data centers and electricity markets. Further, we introduce a benchmark model where data centers have access to the full information about the power system and can identify optimal shifts for the current time period. We demonstrate the efficacy of our model on the IEEE RTS-GMLC system using 5 min load and generation data for an entire year. Our results show that the proposed improvements for the shifting model based on λ CO 2 are highly effective, leading to results that outperform the benchmark model. • Develop a load shifting model to reduce carbon emissions for data center loads. • Introduce a regularization term and cumulative load shifts to existing framework. • Benchmark against a bilevel formulation that enforces DC OPF constraints.

Coordinated Planning of Power Grids in Information Parks considering the Time-Space Transfer Characteristics of Data Center Loads

With the development of green energy, the penetration rate of new energy in the power grid is rising rapidly. Incorporating a large amount of new energy into the power grid will affect the power supply balance of the power grid, and even in severe cases will affect the safe and stable operation of the power grid. The information park has various small and medium data centers. The adjustable load of the data center can be transferred in time and space through the optical fiber and other communication networks, which can enhance the controllability of the power grid to a certain extent and promote the consumption of new energy. This paper proposes a collaborative planning method for information parks that considers the temporal and spatial transfer characteristics of data center loads, builds a load transfer model for data centers and builds a park power grid expansion model with the goal of minimizing the total cost of construction and operation. In this paper, an example of a park is used to verify the calculation example, and it is compared with other schemes to verify the feasibility of the method in this paper.

Dynamic Model and Converter-Based Emulator of a Data Center Power Distribution System

Data centers have become a widespread power electronics (PE) load, which has significant impact on the power grid. In order to investigate the data center load characteristics, this article proposes a complete dynamic model for a typical data center ac power distribution system. A generalized model with mode transition is proposed to coordinate different power stages in the data center power system. Meanwhile, to help evaluate the grid dynamic performance and transient stability, an all-in-one load data center power emulator is developed on a reconfigurable PE converter-based hardware testbed (HTB). The dynamic power model is digitized and simplified to be implemented in two local voltage source inverters on the HTB. This proposed data center power emulator has been verified experimentally in a regional network. Dynamic performances during voltage sag events and server load variations are emulated and discussed. The article details the design, development, and verification of the data center model and power emulator. The proposed model and emulator provide an effective, easy-to-use tool to better design data centers and study the interaction with the power system.

Proliferation of Small Data Networks for Aggregated Demand Response in Electricity Markets

This paper proposes an aggregation method for proliferated small-size data networks to apply data networks’ spatial load regulation potentials for demand response (DR). First, each data network is modeled as a virtual power network, where the Internet data center load is modeled as a set of linear constraints. Second, an aggregated virtual power network (AVPN) is proposed to demonstrate the aggregation method potentials in power systems and to emulate DR applications for multiple virtual power networks in wholesale markets. The coupling of AVPN and power network would develop a linear load model for an aggregated Internet data center. Furthermore, the supply curves representing AVPN DR are formulated to capture heterogeneous regulation costs of virtual power networks, which guarantees limited welfare losses in AVPN DR compared with the individual DR of virtual power networks. Last, an AVPN DR implementation mechanism is deduced to reveal the potentials of the aggregation method in power system applications. Simulation results verify the efficiency of the proposed aggregation method for spatially-coupled DR resources, where, the computing time of the AVPN-based OPF without welfare loss is reduced by 99.94% when there are 3×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</sup> data networks. The proposed aggregation strategy implies that the proliferation of small-size data networks will offer a reasonable DR for enhancing the power system operation in wholesale electricity markets.

Empirical analysis of dynamic load balancing techniques in cloud computing.

Virtualization, dispersed registration, systems administration, programming, and web administrations are all examples of "distributed computing." Customers, datacenters, and scattered servers are just a few of the components that make up a cloud. It includes things like internal failure adaption, high accessibility, flexibility, adaptability, lower client overhead, lower ownership costs, on-demand advantages, and so on. The basis of a feasible load adjusting computation is key to resolving these challenges. CPU load, memory limit, deferral, and system load are all examples of heaps. Burden adjustment is a method for distributing the load across the many hubs of a conveyance framework in order to optimize asset utilization and employment response time while avoiding a situation where some hubs are heavily loaded while others are idle or performing little work. Burden adjustment ensures that at any one time, each processor in the framework or each hub in the system does about the same amount of work. This method may be initiated by the sender, the collector, or the symmetric sort (the blend of sender-started and recipient started types). With some example data center loads, the goal is to create several dynamic load balancing techniques such as Round Robin, Throttled, Equally Spread Current Execution Load, and Shortest Job First algorithms.

Optimal Switching and Load Distribution Strategy for Energy Minimization and Performance Optimization of Cloud Data Center

Cloud computing is a promising computing technology utilized in every stage of the business. The cloud offers different services to cloud users from anytime to anywhere, and it is attained with different parameters, like load optimization, resource optimization. Due to the increase in data center, energy consumption has become a major issue in green data centers. The majority of data centers are function using peak load with huge scales. Thus, it is essential for carrying out energy saving in cloud data centers. This paper designed an energy-saving method using fat tree. The proposed techniques optimize the load at different zones of data center and user in the cloud platform. Here, the distribution of load in cloud data centers is performed using Taylor-based Manta Ray Foraging Optimization (Taylor-MRFO), which is an integration of Manta Ray Foraging Optimization (MRFO) and Taylor series. The method utilized different objectives that involve power, load, latency, and bandwidth. With the load distribution, the switching of cloud data center to the desired mode is performed using Actor critic neural network (ACNN). Thus, the dual strategy leads to performance optimization in cloud infrastructure and also in consolidating parallel workload in data centers more effectively. The proposed Taylor-MRFO+ACNN outperformed other methods with minimal energy of 0.553, minimal load of 0.363, and minimal fitness of 0.437, respectively.

An integrated decision-making framework for sustainable data center operation through intelligent load scheduling

Intelligent load scheduling is an emerging approach that has the potential to facilitate extreme sustainable data center (DC) operation. However, scarcity of straightforward tools in the public domain challenges decision makers performing quantitative analysis of the DC load planning and its potential benefits. In this work, a novel integrated decision-making framework was developed to address this issue, which provides the basis for the multi-objective optimization of carbon-, water-, and economic-intelligent load scheduling. The proposed framework was demonstrated with a case study DC in California, which showed the usefulness of the proposed framework in informing sustainable DC operations.

Study of annual performance and capacity of data center passive cooling mode

This article studies a new architecture of a thermosyphon loop and its capacity to cool data center considering the climate of several cities in France. A parametric experimental study is carried out to show the effect of the refrigerant fill ratio, the input heat load, and the outside temperature. The optimal fill ratio is 16% and the maximum cooling capacity is 1900 W when the data center outside temperature is 20°C. A hydraulic and thermal analytical model is developed to simulate the thermosyphon loop, and it is validated with the experimental results. The cooling load of a typical data center with ten racks, each one dissipates 4.5 kW, is simulated by TRNSYS software for the eight chosen cities in France. The annual performance shows that the percentage of undissipated heat is 13.4% in Marseille that has the highest annual average outside temperature and 2.7% in Caen that has the lowest summer average outside temperature of the data center. The particularity of this work is that it gives the annual performance of a new architecture of a thermosyphon loop in several cities of France with different climates and without electricity consumption.

Adjusting Switching Granularity of Load Balancing for Heterogeneous Datacenter Traffic

The state-of-the-art datacenter load balancing designs commonly optimize bisection bandwidth with homogeneous switching granularity. Their performances surprisingly degrade under mixed traffic containing both short and long flows. Specifically, the short flows suffer from long-tailed delay, while the throughputs of long flows also degrade dramatically due to low link utilization and packet reordering. To solve these problems, we design a traffic-aware load balancing (TLB) scheme to adaptively adjust the switching granularity of long flows according to the load strength of short ones. Under the heavy load of short flows, the long flows use large switching granularity to help short ones obtain more opportunities in choosing short queues to complete quickly. On the contrary, the long flows reroute flexibly with small switching granularity to achieve high throughput. Furthermore, under extremely bursty scenario, we utilize the packet slicing scheme for long flows to release bandwidth for short ones. The experimental results of NS2 simulation and testbed implementation show that TLB significantly reduces the average flow completion time of short flows by 16%-67% over the state-of-the-art load balancers and achieves the high throughput for long flows. Moreover, for extreme bursty case, at the acceptable throughput degradation of long flows, TLB with packet slicing reduces the deadline missing ratio of bursty short flows by up to 80%.

Load balancing with traffic isolation in data center networks

The topologies of current data center networks are typically multi-rooted trees (e.g. leaf–spine) with rich parallel paths between any pair of hosts. Recent progress has demonstrated that effective load balancing can fully utilize these parallel paths to speed up data transfer. Nonetheless, the existing load balancing designs are agnostic to the heterogeneous datacenter traffic, i.e., a mass of delay-sensitive short flows mix with a handful of throughput-oriented long flows, and casually reroute these flows regardless of path condition, thus resulting in frequent flow collisions. The short flows suffer from the problem of large queuing delay due to colliding with long flows, while the frequent collisions between long flows lead to low network utilization and serious throughput degradation. To address these inefficiencies, we propose an isolation-based load balancing scheme, namely ILB, which perceives flow collisions and dynamically isolates the long flows from short flows. Specifically, when a long flow collides with short flows in the same path, it immediately switches to another path unused by short flows to help the short ones in the previous path complete quickly. When short flows disappear, the long flow quickly occupies all its available paths to achieve high throughput. Moreover, ILB is only deployed on the leaf switch without modifying the end-host and spine switch. Experimental results of NS3 simulations show that ILB respectively reduces the average and tail flow completion time for short flows by up to 30% and 70%, as well as increases the throughput of long flows by about 1.68x over the state-of-the-art datacenter load balancing schemes.