Server Racks Research Articles

Investigating thermal performance and energy efficiency in under-floor air distribution data centers: A case study

This study conducted on-site measurements of air temperature and airflow velocity at the cold aisle outlet, combined with infrared imaging, to derive optimized layouts for rack servers. It was found that installing blind plates effectively improves the uniformity of cold air distribution within the racks, as indicated by the temperature and airflow velocity at the server inlet and outlet under both blind and non-blind plate conditions. Finally, thermal performance indicators (i.e., SHI, β, RHI, RCI) and Reynolds number were used to assess the thermal performance of the data center, while energy efficiency indicators (i.e., PUE, WUE, CUE) were employed to evaluate its energy efficiency. The results revealed that the cold aisle has less air distribution at both ends and is more concentrated in the middle. It is advisable to position high-power servers in the middle racks and place racks with relatively lower power at the front and rear ends to ensure effective heat dissipation of the servers. The installation of blind plates can effectively enhance the uniformity of airflow within the racks and improve server cooling. Among the tested racks, 16 out of 17 exhibited low temperatures (RCILO) below 90%, indicating heat loss and energy waste. The PUE value of the data center is 1.19, and the CUE value is 0.77.

Experimental study of an underground pipe gallery pre-cooled chimney ventilation system for a computer room

Natural ventilation is crucial for reducing the energy consumption and carbon emissions of buildings. However, it is often impossible to meet indoor thermal requirements in hot climates. In this study, an integrated natural ventilation system with an underground pipe gallery for pre-cooling and a built-in chimney to enhance the thermal pressure was proposed to cool a computer room. A full-scale experimental platform was constructed to confirm the feasibility of the system. The system performance was investigated by comparing two experimental scenarios, thermal pressure-driven natural ventilation and fan-assisted ventilation, during the transition season in a hot outdoor environment. The underground pipe gallery pre-cooled chimney ventilation system effectively provided a suitable indoor environment for the occupied area, with a mean temperature of 29.1 °C and relative humidity of 57.6 %. The proportion of hours in which the hourly average temperature around the server rack meets the Class C level for computer rooms specified in the national standard is 66.7 %. The maximum pre-cooling capacity of the underground pipe gallery was 3.5 kW. When the assisted fan was operated, the environmental stability of the equipment area further improved, and the system had the potential to operate continuously for 24h to provide a suitable indoor environment.This study provides a strategic solution to passive building design by demonstrating the viability of using an underground pipe gallery for pre-cooling in natural ventilation systems.

Optimization of dielectric oils for immersion liquid cooling of server rack

Abstract In this research, two-phase immersion cooling is utilized to cool a server rack of 8 kW that contains 42 servers. Three-dimensional numerical simulation of server rack model is analyzed by ANSYS 18.1, Computational Fluid Dynamics (CFD). Comparative analysis between two types of dielectric engineered oils is performed, FC-72 and a new oil technology SF-33. The simulation shows a great decrease in temperature ranges for model using SF-33 oil with 34% decrease of the minimum temperature where the most surfaces of servers are at temperatures between 37 °C to 48 °C. Using SF-33 oil makes the hot spot temperature is higher than that for FC-72 by about 3°C. Power consumption for cooling with SF-33 oil is higher due to higher amount of vapor.

Estimation of natural convection heat transfer characteristics of rack server in a cavity: experimental and numerical analyzes

Estimation of natural convection heat transfer characteristics of rack server in a cavity: experimental and numerical analyzes

Dynamic Knowledge Management in an Agent-Based Extended Green Cloud Simulator

Cloud infrastructures operate in highly dynamic environments, and today, energy-focused optimization become crucial. Moreover, the concept of extended cloud infrastructure, which, among others, uses green energy, started to gain traction. This introduces a new level of dynamicity to the ecosystem, as “processing components” may “disappear” and “come back”, specifically in scenarios where the lack/return of green energy leads to shutting down/booting back servers at a given location. Considered use cases may involve introducing new types of resources (e.g., adding containers with server racks with “next-generation processors”). All such situations require the dynamic adaptation of “system knowledge”, i.e., runtime system adaptation. In this context, an agent-based digital twin of the extended green cloud infrastructure is proposed. Here, knowledge management is facilitated with an explainable Rule-Based Expert System, combined with Expression Languages. The tests were run using Extended Green Cloud Simulator, which allows the modelling of cloud infrastructures powered (partially) by renewable energy sources. Specifically, the work describes scenarios in which: (1) a new hardware resource is introduced in the system; (2) the system component changes its resource; and (3) system user changes energy-related preferences. The case study demonstrates how rules can facilitate control of energy efficiency with an example of an adaptable compromise between pricing and energy consumption.

Reliability analysis of CRAC system of a modern Data centre via Kriging algorithm

In the modern era of digital growth, Data centres serve as the important infrastructure of the interconnected global network. The Data centres perform as central hubs where immense volumes of data are processed, stored, and distributed. Reliable performance of data centres and Computer Room Air-Conditioning (CRAC) system is of critical importance, through which optimal environmental conditions are achieved. This paper proposes an interesting approach to evaluate the CRAC system’s ability to function properly under various uncertainties. Heat transfer model is developed using various environmental parameters, data server rack parameters, and evolving capacity of data servers. The uncertainties are modelled as random variables. One of the challenging issues in estimating the reliability of a system under uncertainties is to reduce the computational requirements while maintaining the accuracy. To overcome the issue, Kriging model is developed using adaptive sampling. The proposed approach is compared with the available approaches. The method shows better computational efficiency and accuracy.

TADRP: Toward Thermal-Aware Data Replica Placement in Data-Intensive Data Centers

With the mushrooming growth of data volumes, data replica placement plays a key role in promoting the energy efficiency and Quality-of-Service (QoS) of data-intensive data centers. The existing data placement strategies mainly focus on storage performance improvement or QoS enhancement in data centers, but ignore the indispensable factor -heat recirculation. To bridge this gap, we propose a thermal-aware data replica placement strategy called TADRP, aiming to improve cooling efficiency and minimize the total power consumption of data-intensive data centers. TADRP leverages an ant colony optimization (ACO) algorithm coupled with Laplacian probability distribution to find a quasi-optimal disk sequence (or Disk Sequence for short), which consists of disks selected from different rack servers to place data replicas. TADRP categorizes disks of Disk Sequence into active and inactive ones, by placing hot and cold replicas on active and inactive disks, respectively. We quantitatively evaluate TADRP in terms of cooling costs, total power consumption, number of power-state transitions, and execution time. We compare TADRP with four alternative solutions, namely, Random, Hadoop, SRS, and CDP-NSGAII. Experimental results show that TADRP can reduce the cooling costs and the total power of the existing solutions by 14.7% -61.7% and 19.2 %-55.1%, respectively, without undesirable I/O performance drops.

Design and analysis of wireless data center network topology HCDCN based on VLC

With the rapid development of the Internet of Things, big data, cloud computing and other industries, the data center network, as the core infrastructure of the computing industry, is becoming increasingly important. Wireless communication technology is developing rapidly. This paper uses visible light wireless communication technology to build a wireless data center network, solving the problems of complex wiring and expansion difficulties in the existing data center network.Aiming at the problems of insufficient spectrum, low bandwidth utilization and large energy consumption in the existing wireless data center, this paper adopts the visible light wireless communication technology based on LED, and proposes a new wireless data center network structure HCDCN. Firstly, work have been done to analyze the feasibility of introducing data centers in this technology, and transform server racks in combination with the characteristics of the technology. Secondly, the construction process of the structure is introduced and a new coding rule is proposed. In order to increase the routing efficiency and fault tolerance, the rack top link routing path is added, and an efficient routing algorithm is designed. Finally, the aggregate throughput of HCDCN under different scales is obtained through experiments, and the network performance of HCDCN structure is compared with that of OWCells and Mesh structure. Experiments have proved that HCDCN's network performance is better than the other two structures under the same conditions. The new wireless data center network proposed in this paper has largely alleviated the bandwidth shortage of hotspot servers.

Modeling Convective Heat Transfer of Air in a Data Center Using OpenFOAM

Achieving energy and cooling efficiency in data center convective air flow and heat transfer can be a challenging task. Among different numerical methods to work with such issues is the Finite Volume Method in Computational Fluid Dynamics. This work evaluates the performance of two such solvers provided by OpenFOAM® in solving this type of convective heat-transfer problem, namely BuoyantBoussinesqPimpleFOAM and BuoyantPimpleFOAM. This is done for two different flow configurations of significantly different Richardson number. To sufficiently resolve the flow, grid sizing effects are elucidated by way of the kernel density estimate. It determines the volume distribution of the temperature in the data center configuration. For the k-epsilon turbulence model used here, it was found that the compressible solver performs faster and requires less grid resolution for both flow configurations. This is attributed to the nature of the boundary conditions which are set such that the mass flow conservation per server rack and cooling unit is achieved. Transient solutions are found to provide better iterative convergence for cases that involves buoyancy, compressibility and flow separation. This is, in comparison to steady-state solutions where artificial numerical pressure drop is found, to depend on the momentum relaxation factors for the convective case with a higher Richardson number.

Modeling and Thermal Simulation of 2U-Rack Servers Applying Computational Fluid Dynamics Techniques

The data center industry consumes between 196 and 400 terawatt-hours annually, between 1% and 2% of the world's energy consumption. A high percentage of data center energy consumption is associated with air conditioning and cooling systems. The optimization of the data center cooling process can occur at multiple scales, being one of the most relevant to the generation and heat transfer processes within servers. In this approach, we must study the temperature fields of each server’s components and understand airflow behavior within the server chassis. The present work, in its first stage, performs a thermal and cooling analysis for a two-unit rack server (2U-Rack) using computational fluid dynamics (CFD) techniques via Ansys Fluent code. A geometric server model was developed, considering heat generation in the main electrical components and heat and air mass transfer with the data center. The model was validated by comparing simulation results with the server's sensor values and the results of thermography testing. Following CFD simulation, airflow velocity fields, temperature contours, and sensitivity scenarios were obtained, with the goal of proposing an optimization model that would allow us to improve server heat transfer processes in the following steps.

4 Ways to Put Lasers on Silicon: You Can Make Many Things with Silicon Photonics, But a Laser is not One of Them

Photonic Integrate Circuita, which combine a collection of optoelectronic functions on a single chip, are an increasingly common part of everyday life. They are used in high-speed optical transceivers that link server racks in data centers, including the one used to deliver the IEEE Spectrum website, in lidars to keep self-driving cars on track, and in spectrometers to spot chemicals in the atmosphere, among many other applications. All these systems have grown less expensive and, in some cases have become economically feasible, by making most of the IC with silicon fabrication technologies.

Impedance-Based Aggregation of Paralleled Power Factor Correction Converters in Data Centers

Existing aggregation methods for large-scale data center power systems often overlook disparities in power supply units, leading to reduced accuracy in the small signal dynamics of aggregated model. To fill this gap, this article develops an impedance-based aggregation method for paralleled power factor correction (PFC) converters with heterogeneities in power levels, conduction modes, and line impedances. The closed-loop input admittances of PFC converters operating in continuous conduction mode (CCM) and discontinuous conduction mode (DCM) are presented first. The parametric scaling is then derived to aggregate multiple PFC converters at different power levels and operation modes, including the CCM, DCM, and mixed conduction mode, at the rack level. Furthermore, the aggregation rule and criteria for paralleled PFC converters with unequal line impedances is developed to provide a reduced-order dynamic model for adjacent server racks. The proposed parameter scaling and aggregation method allow efficient interaction analysis and stability assessment for data center power systems. The accuracy and robustness of the impedance-based aggregation are finally validated by simulations and experimental tests.

A two-phase heuristic algorithm for power-aware offline scheduling in IaaS clouds

The paper aims at mitigating hot-spots during Offline Scheduling in IaaS (Infrastructure-as-a-Service) cloud systems. Unlike previous studies, the research focuses on identifying and resolving hot-spots not at servers, but at server racks. A two-phase algorithm for performing power-aware offline scheduling is proposed. The first phase aims at identifying and mitigating hot-spots at racks, while the second phase performs VM consolidation, i.e. minimization of the number of occupied servers while maintaining a feasible VM mapping and low migration costs. The proposed algorithm takes into account the dynamic nature of VM's resource consumption: it does not only resolve detected hot-spots, but also tries to avoid hot-spots in a reasonable future time period. The algorithm was tested with the data from a real IaaS cloud with different sets of algorithm's parameters. Experimental evaluation showed that the statistical estimates of the future VM's resource consumption provide the most reliable mapping, which is a result of minimization of the number of new hot-spot occurrences.

Numerical study on the optimal power distribution of server racks in a data center

Numerical study on the optimal power distribution of server racks in a data center

A data-driven subspace predictive control method for air-cooled data center thermal modelling and optimization

This paper presents a data-driven predictive control method for optimizing the energy consumption of air-cooled data centers with unknown system model parameters. First, based on the measurable data of the studied system, the subspace predictive control (SPC) method is adopted to improve the energy use efficiency of the data center by regulating the power allocation of the server racks and the supply temperature of cold air, while ensuring the safe operating environment of the electronic equipment. Furthermore, a reasonable event-triggered law is designed to solve the problem of the low computational efficiency of the conventional SPC method. The simulation results illustrate that the designed event-triggered law can improve the computational efficiency of the algorithm while maintaining the control performance of the algorithm, which verifies its application prospect in practice.

Fan-noise reduction of data centre telecommunications’ server racks, with an acoustic metamaterial broadband, low-frequency sound-absorbing liner

An acoustic metamaterial is evaluated successfully as a means to reduce axial fan noise in an IT server rack application. The continuing growth in demand for cloud-based memory storage is outpacing sustainable energy management which still requires the axial fan as a fundamental component in temperature control technologies. As chip temperatures increase, additional numbers of more powerful fans are required to run faster with the result that data centres have become unsafe working environments due to the acoustic output exceeding prescribed occupational noise level limits. Noise isolation using soundproof cabinets is expensive and impractical in data centres which seek to reduce floor space and communication losses. Instead, in this work, a 15 mm deep acoustic liner solution is proposed being the approach taken in the aeronautics sector to reduce aeroengine noise which, similarly, is a flow duct problem. The acoustic metamaterial adopted here is the recently developed Segmented Membrane Sound Absorber (SeMSA) which can be tailored for low frequency, broadband and tonal noise in a thin form-factor and which can be used in a grazing flow configuration. An equivalent circuit model developed for the SeMSA allows its design to be optimised to attenuate a typical server fan’s frequency spectrum. An array of manufactured rectangular SeMSA cells was embedded in a “deflector plate”, which is an existing component in a server rack. The deflector plate was implemented in a custom-built fan sound power rig, tested in an hemi-anechoic chamber and compared to a hardwall case and to PU foam. A relative overall decrease of 2.5 dBA was achieved for the SeMSA deflector plate liner which corresponds to a 43.9% decrease in overall sound emissions, which using a material thickness of only 15 mm, signifies an almost twofold increase in the time required for hearing damage to occur. A 3 dBA drop was achieved at the blade passing frequency of 620 Hz, while relative drops of 1–2.5 dBA were achieved from 2000–8000 Hz.

A Multi-Port Hardware Energy Meter System for Data Centers and Server Farms Monitoring.

Nowadays the rationalization of electrical energy consumption is a serious concern worldwide. Energy consumption reduction and energy efficiency appear to be the two paths to addressing this target. To achieve this goal, many different techniques are promoted, among them, the integration of (artificial) intelligence in the energy workflow is gaining importance. All these approaches have a common need: data. Data that should be collected and provided in a reliable, accurate, secure, and efficient way. For this purpose, sensing technologies that enable ubiquitous data acquisition and the new communication infrastructure that ensure low latency and high density are the key. This article presents a sensing solution devoted to the precise gathering of energy parameters such as voltage, current, active power, and power factor for server farms and datacenters, computing infrastructures that are growing meaningfully to meet the demand for network applications. The designed system enables disaggregated acquisition of energy data from a large number of devices and characterization of their consumption behavior, both in real time. In this work, the creation of a complete multiport power meter system is detailed. The study reports all the steps needed to create the prototype, from the analysis of electronic components, the selection of sensors, the design of the Printed Circuit Board (PCB), the configuration and calibration of the hardware and embedded system, and the implementation of the software layer. The power meter application is geared toward data centers and server farms and has been tested by connecting it to a laboratory server rack, although its designs can be easily adapted to other scenarios where gathering the energy consumption information was needed. The novelty of the system is based on high scalability built upon two factors. Firstly, the one-on-one approach followed to acquire the data from each power source, even if they belong to the same physical equipment, so the system can correlate extremely well the execution of processes with the energy data. Thus, the potential of data to develop tailored solutions rises. Second, the use of temporal multiplexing to keep the real-time data delivery even for a very high number of sources. All these ensure compatibility with standard IoT networks and applications, as the data markup language is used (enabling database storage and computing system processing) and the interconnection is done by well-known protocols.

Tuning block-parallel all-pairs shortest path algorithm for efficient multi-core implementa- tion.

Finding shortest paths in a weighted graph is one of the key problems in computer-science, which has numerous practical applications in multiple domains. This paper analyzes the parallel blocked all-pairs shortest path algorithm at the aim of evaluating the influence of the multi-core system and its hierarchical cache memory on the parameters of algorithm implementation depending on the size of the graph and the size of distance matrix’s block. It proposes a technique of tuning the block-size to the given multi-core system. The technique involves profiling tools in the tuning process and allows the increase of the parallel algorithm throughput. Computational experiments carried out on a rack server equipped with two Intel Xeon E5-2620 v4 processors of 8 cores and 16 hardware threads each have convincingly shown for various graph sizes that the behavior and parameters of the hierarchical cache memory operation don’t depend on the graph size and are determined only by the distance matrix’s block size. To tune the algorithm to the target multi-core system, the preferable block size can be found once for the graph size whose in-memory matrix representation is larger than the size of cache shared among all processor’s cores. Then this blocksize can be reused on graphs of bigger size for efficient solving the all-pairs shortest path problem

Malcolm: Multi-agent Learning for Cooperative Load Management at Rack Scale

We consider the problem of balancing the load among servers in dense racks for microsecond-scale workloads. To balance the load in such settings tens of millions of scheduling decisions have to be made per second. Achieving this throughput while providing microsecond-scale latency and high availability is extremely challenging. To address this challenge, we design a fully decentralized load-balancing framework. In this framework, servers collectively balance the load in the system. We model the interactions among servers as a cooperative stochastic game. To find the game's parametric Nash equilibrium, we design and implement a decentralized algorithm based on multi-agent-learning theory. We empirically show that our proposed algorithm is adaptive and scalable while outperforming state-of-the art alternatives. In homogeneous settings, Malcolm performs as well as the best alternative among other baselines. In heterogeneous settings, compared to other baselines, for lower loads, Malcolm improves tail latency by up to a factor of four. And for the same tail latency, Malcolm achieves up to 60% more throughput compared to the best alternative among other baselines.

Thermal Performance of the Thin Heat Pipe for Cooling of Solid-State Drives

With the rapid development of information science and technology, the demand for computer data processing is increasing, resulting in the rapid growth of the demand for high-power and high-performance solid-state drives (SSDs). The stable operation of SSDs plays an important role in ensuring the reliable working conditions and appropriate temperature of information technology equipment, rack servers, and related facilities. However, SSDs usually have significant heat emissions, putting forward higher requirements for temperature and humidity control, and consequently the heat sink system for cooling is essential to maintain the proper working state of SSDs. In this paper, a new type of thin heat pipe (THP) heat sink is proposed, and the heat transfer performance and cooling effect are experimentally and numerically studied. The numerical results are compared with experimental results, which showed an error within 5%. Single and double heat pipes were investigated under different input powers (from 5 W to 50 W) and different placement angles between 0° and 90°. The heat transfer performance of the new heat sink is analyzed by the startup performance, the evaporator temperature, and the total thermal resistance. The results show that the new double THPs with a 90° angle have a great advantage in the heat transfer performance of SSDs. The research is of great significance for the design and optimization of the SSDs’ cooling system in practical applications.