Commercial Data Centers Research Articles

An Exploratory Analysis of the Commercial Data Center Location in the Republic of Korea for dispersing Non-capital Region

An Exploratory Analysis of the Commercial Data Center Location in the Republic of Korea for dispersing Non-capital Region

Analysis of Anthropogenic Waste Heat Emission from an Academic Data Center

The rapid growth in computing and data transmission has significant energy and environmental implications. While there is considerable interest in waste heat emission and reuse in commercial data centers, opportunities in academic data centers remain largely unexplored. In this study, real-time onsite waste heat data were collected from a typical academic data center and an analysis framework was developed to determine the quality and quantity of waste heat that can be contained for reuse. In the absence of a comprehensive computer room monitoring system, real-time thermal data were collected from the data center using two arrays of thermometers and thermo-anemometers in the server room. Additionally, a computational fluid dynamics model was used to simulate temperature distribution and identify “hot spots” in the server room. By simulating modification of the server room with a hot air containment system, the return air temperature increased from 23 to 46 °C and the annual waste heat energy increased from 377 to 2004 MWh. Our study emphasizes the importance of containing waste heat so that it can be available for reuse, and also, that reusing the waste heat has value in not releasing it to the environment.

Application of SDN-IP hybrid network multicast architecture in Commercial Aerospace Data Center

Abstract The increasing amount of massive data generated by commercial spacecraft in orbit puts forward higher and higher requirements for the stability, reliability, and computing power of computer systems for commercial aerospace data center. Data center computer systems are gradually transforming from X86 architecture and IP network model to a platform model with cloud computing and software-defined network (SDN) technology. This article proposes a new network architecture based on a unicast/multicast protocol for data interaction between the SDN and IP network. There are three main contributions of this article. The first is that the architecture proposed in this article aims to reduce end-to-end transmission latency and packet loss. The second is to improve the flexibility of system configuration and precise control when the SND controller state changes. The third is to verify the feasibility of deploying network architecture in a real data center environment.

An Experimental Study on the Effect of Nanofluids on the Thermal Conductivity and Rheological Properties of a Coolant for Liquids

Thermal conductivity and viscosity are important properties for nanofluids as they significantly affect the flow and heat transfer process. To date, the rheological properties of water-based nanofluids have been well studied, while the results are scarce for non-aqueous nanofluids. In this study, the thermal conductivity and rheological properties of two different kinds of oxide nanofluids (CuO and Al2O3) in a typical commercial data center focusing on liquid coolants were systematically investigated at different mass fractions and temperatures. The results showed that the addition of nanoparticles can significantly improve the heat conduction capacity of mineral oil coolants. There is an average increase in thermal conductivity of up to 20–25%. The shear rate–shear stress and shear rate–viscosity curves all showed that mineral oil coolant-based oxide nanofluids behaved as Newtonian fluids and that nanoparticles did not cause the increment in viscosity. The effect of temperature on rheological properties was also studied, and the result showed that high temperatures resulted in low viscosity and shear stress. Finally, the effect of particle type was investigated, and it was found that no matter what kind of nanoparticles were added, their effects on the rheological behaviors were the same.

Medium- and Long-term Forecast of China’s Electricity Consumption Considering the Fast Growth of New Infrastructure

In order to predict China’s electricity consumption demand under situation of fast growth of the new infrastructure, a medium- and long-term electricity consumption demand prediction model based on LEAP is proposed, which calculates the overall electricity consumption of the country according to the activity level and intensity of typical new infrastructure and other sectors. Five different scenarios are set for comparison. The results show that under the basic scenario, China’s electricity consumption will reach 7408 billion kWh in 2035, in which the consumption of new infrastructure accounts for 31.31%.While under the other four scenarios, China’s electricity consumption will reach to the minimum of 7319 billion kWh under the scenario of improvement of energy-saving technology, and the maximum of 7525 billion kWh under the scenario of fast development of the new infrastructure, with higher contributions of the consumption of commercial charging, data centers and 5G base station. Relevant suggestions are put forward from the aspects of new infrastructure construction, development of clean energy and technical standards to reduce the electricity consumption of new infrastructure.

Prospects for the Development of Data Centers in the Arctic

This paper examines the experience and prospects of placing data centers in the Arctic zone of circumpolar states, including Russia. The authors consider the phenomenon of commercial data centers providing services to IT-companies and the idea of building such centers in the Arctic to reduce costs. The authors analyze the advantages of the Arctic territories, including the possibility of natural cooling of installations, the use of renewable energy sources and the prospects for contribution to digitalization and development of the northern regions of the Arctic countries. Further, the authors describe the existing projects for the placement of data centers in the Arctic zone of Sweden, Finland, Norway, Denmark, Iceland and similar Russian experience. In this paper, the authors dwell in detail on projects in Karelia, Murmansk, and Norilsk. In conclusion, the authors come to the conclusion that functioning data centers are isolated projects both in Russia and in other countries. The authors analyze in detail what features of the Arctic regions determine this state of affairs, both highlighting common features and distinguishing between the Russian and foreign Arctic.

Intelligent and compliant dynamic software license consolidation in cloud environment

Based on the virtualization technology and pushed by the softwarization paradigm and the actual demand for services and resources, commercial cloud data centers know an unprecedented expansion. The systematic presence of software and the services generated enabled the development of the already dense application expenses. This causes, not only a cost explosion, especially when proprietary solutions protected by licenses are in hand, but also, represents a critical need in terms of software asset and resource management at the SaaS level. In addition to these costs, inefficient resource utilization, and the resulting energy represent an important part of the operational expenditure of data centers and are still a hot topic despite the consolidation initiatives put in place. The main objective of the consolidation service is to maximize resource exploitation while minimizing energy consumption and costs, among others. Even so, we have noticed that the reported literature doesn’t treat license management in the cloud environment as a whole, especially, from the resource management perspective and the overwhelming majority of the consolidation work focuses on resource optimization at the IaaS level. Therefore, we propose a reinforcement learning-based scheme that allows efficient use of resources and optimizes costs, energy consumption, and resource wastage, while remaining compliant. The experimental results show that our intelligent consolidator outperforms the baseline approaches according to the evaluation metrics used regardless of the resource heterogeneity and the data center dimensionality.

Line segment visibility with sidedness constraints

We study a family of line segment visibility problems, related to classical art gallery problems, which are motivated by monitoring requirements in commercial data centers. Given a collection of non-overlapping line segments in the interior of a rectangle, and a requirement to monitor the segments from one side or the other, we examine the problem of finding a minimal set of point guards. Guards may be placed anywhere in the interior of the rectangle but not on a line segment. We consider combinatorial bounds of problem variants where the problem solver gets to decide which side of the segments to guard or the problem poser gets to decide which side to guard, and many others. We show that virtually all variants are NP-Hard to solve exactly, and then provide heuristics and experimental results to give insight into the associated practical problems. Finally we describe a program for using experiments to guide the search for optimal combinatorial bounds.

Prospects and Challenges of Photonic Switching in Data Centers and Computing Systems

This Tutorial will discuss the motivation, benefits, and challenges of photonic switching in data centers and cover prospects of future data centers involving emerging new technologies and cross-layer solutions. The primary motivation for considering photonic switching in data centers rises from the need for energy-efficient and scalable intra-data center networks to meet rapid increases in data traffic driven by emerging applications, including machine learning. The data traffic inside the data centers (East-West traffic) is typically significantly greater than that of the traffic coming in and out of the data centers (North-South traffic) (Benson <i>et al.</i> 2010). To accommodate such traffic, today&#x0027;s large-scale data centers employ cascaded stages of many power-hungry electronic packet switches interconnected across the data center network in fixed hierarchical communication topologies. These electronic switches add significant latency and energy consumption while limiting the communication bandwidth. On the other hand, photonic switches can, in principle, support interconnections at very high data rates on many parallel wavelengths while keeping their energy consumption nearly independent of the switch port bandwidth. Numerous research papers have predicted significant benefits in scalability, throughput, and power efficiency from deploying photonic switches in data centers. However, photonic switching is not yet widely deployed in commercial warehouse-scale data centers at the time of writing this Tutorial due to significant challenges. They are related to 1) cross-layer issues involving control and management planes together with data integrity during switching, 2) scalability to &gt;5000 racks (&gt;a quarter-million servers), 3) performance monitoring required for reliable operation, 4) currently existing standards allowing limited power margin (3 dB), and 5) other practical (technology-dependent) issues relating to polarization sensitivity, temperature sensitivity, cost, etc. In telecom, commercial deployments of reconfigurable optical add-drop multiplexers (ROADMs) (Lightwave, 2003)&#x2013;(Perrin, 2015) also had faced similar challenges and took place nearly ten years after the first research network testbed demonstrations in 1997 (Garrett, 2015) with fully implemented network control and management planes. In data centers, the challenges are far more significant due to the scale of the network and the dynamicity of the traffic. We will discuss possible solutions for future data centers involving cross-layer methods, new topologies, and innovative photonic switching technologies. In particular, the Tutorial broadly surveys state-of-the-art photonic switching technologies, architectures, and experimental results, and further covers the details of arrayed-waveguide-grating-router-based switch fabrics offering hybrid switching methods with distributed control planes towards scalable data center networking.

Modeling the Performance of the Clock Phase Caching Approach to Clock and Data Recovery

Optical switching could enable data center networks to keep pace with the rapid growth of intra-data center traffic, however, sub-nanosecond clock and data recovery time is crucial to enabling optically-switched data center networks to transport small packet dominated data center traffic with over 90% efficiency. We review the clock-synchronized approach to clock and data recovery, which enables sub-nanosecond switching time in optically switched networks. We then introduce an analytical model to mathematically explore the operation of clock phase caching, and use this model to explore the impact of factors such as fiber temperature, clock jitter and symbol rate on the BER and clock and data recovery locking time performance of the clock phase caching approach, as well as their impact on scalability. Using commercial data center parameters matching those used in our previous experimental research, we find that our analytical model provides estimates that closely match our previous experimental results, validating its use for making predictions of the performance of clock phase cached systems.

Power-optimized Deployment of Key-value Stores Using Storage Class Memory

High-performance flash-based key-value stores in data-centers utilize large amounts of DRAM to cache hot data. However, motivated by the high cost and power consumption of DRAM, server designs with lower DRAM-per-compute ratio are becoming popular. These low-cost servers enable scale-out services by reducing server workload densities. This results in improvements to overall service reliability, leading to a decrease in the total cost of ownership (TCO) for scalable workloads. Nevertheless, for key-value stores with large memory footprints, these reduced DRAM servers degrade performance due to an increase in both IO utilization and data access latency. In this scenario, a standard practice to improve performance for sharded databases is to reduce the number of shards per machine, which degrades the TCO benefits of reduced DRAM low-cost servers. In this work, we explore a practical solution to improve performance and reduce the costs and power consumption of key-value stores running on DRAM-constrained servers by using Storage Class Memories (SCM). SCMs in a DIMM form factor, although slower than DRAM, are sufficiently faster than flash when serving as a large extension to DRAM. With new technologies like Compute Express Link, we can expand the memory capacity of servers with high bandwidth and low latency connectivity with SCM. In this article, we use Intel Optane PMem 100 Series SCMs (DCPMM) in AppDirect mode to extend the available memory of our existing single-socket platform deployment of RocksDB (one of the largest key-value stores at Meta). We first designed a hybrid cache in RocksDB to harness both DRAM and SCM hierarchically. We then characterized the performance of the hybrid cache for three of the largest RocksDB use cases at Meta (ChatApp, BLOB Metadata, and Hive Cache). Our results demonstrate that we can achieve up to 80% improvement in throughput and 20% improvement in P95 latency over the existing small DRAM single-socket platform, while maintaining a 43–48% cost improvement over our large DRAM dual-socket platform. To the best of our knowledge, this is the first study of the DCPMM platform in a commercial data center.

Food for thought: Commentary on Burnette et al. (2021) "Concerns and recommendations for using Amazon MTurk for eating disorder research".

Burnette et al. aimed to validate two eating disorder symptom measures among transgender adults recruited from Mechanical Turk (MTurk). After identifying several data quality issues, Burnette et al. abandoned this aim and instead documented the issues they faced (e.g., demographic misrepresentation, repeat submissions, inconsistent responses across similar questions, failed attention checks). Consequently, Burnette et al. raised concerns about the use of MTurk for psychological research, particularly in an eating disorder context. However, we believe these claims are overstated because they arise from a single study not designed to test MTurk data quality. Further, despite claiming to go "above and beyond" current recommendations, Burnette et al. missed key screening procedures. In particular, they missed procedures known to prevent participants who use commercial data centers (i.e., server farms) to hide their true IP address and complete multiple surveys for financial gain. In this commentary, we outline key screening procedures that allow researchers to obtain quality MTurk data. We also highlight the importance of balancing efforts to increase data quality with efforts to maintain sample diversity. With appropriate screening procedures, which should be preregistered, MTurk remains a viable participant source that requires further validation in an eating disorder context.

Heuristic-based Approach for Dynamic Consolidation of Software Licenses in Cloud Data Centers

Since its emergence, cloud computing has continued to evolve thanks to its ability to present computing as consumable services paid by use, and the possibilities of resource scaling that it offers according to client’s needs. Models and appropriate schemes for resource scaling through consolidation service have been considerably investigated, mainly, at the infrastructure level to optimize costs and energy consumption. Consolidation efforts at the SaaS level remain very restrained mostly when proprietary software are in hand. In order to fill this gap and provide software licenses elastically regarding the economic and energy-aware considerations in the context of distributed cloud computing systems, this work deals with dynamic software consolidation in commercial cloud data centers 〖DS〗^3 C. Our solution is based on heuristic algorithms and allows reallocating software licenses at runtime by determining the optimal amount of resources required for their execution and freed unused machines. Simulation results showed the efficiency of our solution in terms of energy by 68.85% savings and costs by 80.01% savings. It allowed to free up to 75% physical machines and 76.5% virtual machines and proved its scalability in terms of average execution time while varying the number of software and the number of licenses alternately.

Customized data center cooling system operating at significant outdoor temperature fluctuations

This elaboration presents the configuration of the cooling system of the POLCOM Data Center which utilizes commercially available components, proposes steering strategy and analyzes their functionality. The designed architecture of cooling and dedicated control system is presented to demonstrate the novelty and customization with respect to the demanding outdoor temperatures existing in the climate of the Małopolska Province. The cooling system control and operation is illustrated by a comprehensive analysis of compressor and freecooling modes. The discussion on embedded functionality of chillers is undertaken. According to features and limitations of the complex solution, the cooling system reached the annual average coefficient of performance of 8.63 in 2015 (increase of 110% compared to the year 2014), operating 65% of the time during the year in compressor mode and 35% in freecooling mode. This coefficient in compressor mode amounted to 4.39 while in freecooling mode totaled 16.50. It was proved that in the real case under consideration they generated losses in electricity consumption amounting to 557MWh per year. The real-time experimental data collected from the commercial Data Center installation are used to present a unique operation of a such complex system.

Performance Analyses of Temperature Controls by a Network-Based Learning Controller for an Indoor Space in a Cold Area

For the sustainable use of building spaces, various methods have been studied to satisfy specific conditions required by the characteristics of space types and the energy use in operation. However, several effective control approaches adopting the latest statistical tools may have problems such as higher control precision increases energy consumption, or lower energy consumption decreases their control precision. This study proposes an optimized model to reach the indoor set-point temperature by controlling the amount of heating supply air and its temperature and investigates the efficiency of an adaptive controller to maintain indoor thermal comfort within setting ranges. In the consistency of the comfort level, the fuzzy logic controller was found to be 1.76% and the artificial neural network controller to be 17.83%, respectively, more efficient than the conventional thermostat. In addition, for energy use efficiency, both of the controllers were confirmed to be over 3.0% more efficient. Consequently, the network-based controller with the adaptive controller checking comfort levels effectively works to improve both energy efficiency and thermal comfort. This improvement can be significant in places such as commercial high-rises, large hospitals, and data centers where many spaces are intensively woven with appropriate thermal environments to maintain users’ workability.

A Machine Learning Solution for Data Center Thermal Characteristics Analysis

The energy efficiency of Data Center (DC) operations heavily relies on a DC ambient temperature as well as its IT and cooling systems performance. A reliable and efficient cooling system is necessary to produce a persistent flow of cold air to cool servers that are subjected to constantly increasing computational load due to the advent of smart cloud-based applications. Consequently, the increased demand for computing power will inadvertently increase server waste heat creation in data centers. To improve a DC thermal profile which could undeniably influence energy efficiency and reliability of IT equipment, it is imperative to explore the thermal characteristics analysis of an IT room. This work encompasses the employment of an unsupervised machine learning technique for uncovering weaknesses of a DC cooling system based on real DC monitoring thermal data. The findings of the analysis result in the identification of areas for thermal management and cooling improvement that further feeds into DC recommendations. With the aim to identify overheated zones in a DC IT room and corresponding servers, we applied analyzed thermal characteristics of the IT room. Experimental dataset includes measurements of ambient air temperature in the hot aisle of the IT room in ENEA Portici research center hosting the CRESCO6 computing cluster. We use machine learning clustering techniques to identify overheated locations and categorize computing nodes based on surrounding air temperature ranges abstracted from the data. This work employs the principles and approaches replicable for the analysis of thermal characteristics of any DC, thereby fostering transferability. This paper demonstrates how best practices and guidelines could be applied for thermal analysis and profiling of a commercial DC based on real thermal monitoring data.

Feature Selection for Improving Failure Detection in Hard Disk Drives Using a Genetic Algorithm and Significance Scores

Hard disk drives (HDD) are used for data storage in personal computing platforms as well as commercial datacenters. An abrupt failure of these devices may result in an irreversible loss of critical data. Most HDD use self-monitoring, analysis, and reporting technology (SMART), and record different performance parameters to assess their own health. However, not all SMART attributes are effective at detecting a failing HDD. In this paper, a two-tier approach is presented to select the most effective precursors for a failing HDD. In the first tier, a genetic algorithm (GA) is used to select a subset of SMART attributes that lead to easily distinguishable and well clustered feature vectors in the selected subset. The GA finds the optimal feature subset by evaluating only combinations of SMART attributes, while ignoring their individual fitness. A second tier is proposed to filter the features selected using the GA by evaluating each feature independently, using a significance score that measures the statistical contribution of a feature towards disk failures. The resultant subset of selected SMART attributes is used to train a generative classifier, the naïve Bayes classifier. The proposed method is tested on a SMART dataset from a commercial datacenter, and the results are compared with state-of-the-art methods, indicating that the proposed method has a better failure detection rate and a reasonable false alarm rate. It uses fewer SMART attributes, which reduces the required training time for the classifier and does not require tuning any parameters or thresholds.

A Systems Overview of Commercial Data Centers

Data center facilities play a vital role in present and forthcoming information and communication technologies. Internet giants, such as IBM, Microsoft, Google, Yahoo, and Amazon hold large data centers to provide cloud computing services and web hosting applications. Due to rapid growth in data center size and complexity, it is essential to highlight important design aspects and challenges of data centers. This article presents market segmentation of the leading data center operators and discusses the infrastructural considerations, namely energy consumption, power usage effectiveness, cost structure, and system reliability constraints. Moreover, it presents data center network design, classification of the data center servers, recent developments, and future trends of the data center industry. Furthermore, the emerging paradigm of mobile cloud computing is debated with respect to the research issues. Preliminary results for the energy consumption of task scheduling techniques are also provided.

IBM z14: Improved datacenter characteristics, energy efficiency, and packaging innovation

The IBM z14 mainframe provides significant improvements in datacenter physical characteristics, flexibility, robustness, energy efficiency, and modeling capability relative to its predecessor, the IBM z13. In this paper, we consider these improvements in the context of the latest trends in the commercial datacenter and describe the innovations in packaging, controls, testing, modeling, and tools that enabled them. The z14 supports the American Society of Heating Refrigeration and Air Conditioning Engineers class A3 maximum air-inlet temperature of 40 °C, versus 35 °C for the IBM z13. Water-cooled models have increased maximum inlet water temperature. A maximum-configuration IBM z14 delivers 23% or more traditional compute capacity per watt than the z13. The z14 offers a cover set option that has a total system depth of 1.486 m versus 1.87 m for z13, which is much closer to the 1.20-m depth of many datacenter racks. The z14 has a high level of physical robustness with respect to shipping and operating shock and vibration, as well as earthquakes. We discuss how this robustness was achieved through design, modeling, and testing. We also review the capabilities of tooling to accurately estimate system power, weight, and airflow, as well as the heat extracted to water in the water-cooled systems.

Scalable topological forwarding and routing policies in RINA‐enabled programmable data centers

AbstractGiven the current expansion of cloud computing, the expected advent of the Internet of Things, and the requirements of future fifth‐generation network infrastructures, significantly larger pools of computational and storage resources will soon be required. This emphasizes the need for more scalable data centers that are capable of providing such an amount of resources in a cost‐effective way. A quick look into today's commercial data centers shows that they tend to rely on variations of well‐defined leaf‐spine/Clos data center network (DCN) topologies, offering low latency, ultrahigh bisectional bandwidth, and enhanced reliability against concurrent failures. However, DCNs are typically restricted by the use of the Transmission Control Protocol/Internet Protocol (TCP/IP) suite, thus suffering limited routing scalability. In this work, we study the benefits that replacing TCP/IP with the recursive internetwork architecture (RINA) can bring into commercial DCNs, focusing on forwarding and routing scalability. We quantitatively evaluate the benefits that RINA solutions can yield against those based on TCP/IP and highlight how, by deploying RINA, topological routing solutions can improve even more the efficiency of the network. To this goal, we propose a rule‐and‐exception forwarding policy tailored to the characteristics of several DCN variants, enabling fast forwarding decisions with merely neighbors' information. Upon failures, few exceptions are necessary, whose computation can also profit from the known topology. Extensive numerical results show that the proposed policy requirements depend mainly on the number of neighbors and concurrent failures in the DCN rather than its size, dramatically reducing the amount of forwarding and routing information stored at DCN nodes.