Data Center Power Research Articles

Proactive Demand Response for Data Centers: A Win-Win Solution

In order to reduce the energy cost of data centers, recent studies suggest distributing computation workload among multiple geographically dispersed data centers, by exploiting the electricity price difference. However, the impact of data center load redistribution on the power grid is not well understood yet. This paper takes the first step towards tackling this important issue, by studying how the power grid can take advantage of the data centers' load distribution proactively for the purpose of power load balancing. We model the interactions between power grid and data centers as a two-stage problem, where the utility company chooses proper pricing mechanisms to balance the electric power load in the first stage, and the data centers seek to minimize their total energy cost by responding to the prices in the second stage. We show that the two-stage problem is a bilevel quadratic program, which is NP-hard and cannot be solved using standard convex optimization techniques. We introduce benchmark problems to derive upper and lower bounds for the solution of the two-stage problem. We further propose a branch and bound algorithm to attain the globally optimal solution, and propose a heuristic algorithm with low computational complexity to obtain an alternative close-to-optimal solution. We also study the impact of background load prediction error using the theoretical framework of robust optimization. The simulation results demonstrate that our proposed scheme can not only improve the power grid reliability but also reduce the energy cost of data centers.

Coordinating workload balancing and power switching in renewable energy powered data center

There has been growing concern about energy consumption and environmental impact of datacenters. Some pioneers begin to power datacenters with renewable energy to offset carbon footprint. However, it is challenging to integrate intermittent renewable energy into datacenter power system. Grid-tied system is widely deployed in renewable energy powered datacenters. But the drawbacks (e.g. Harmonic disturbance and costliness) of grid tie inverter harass this design. Besides, the mixture of green load and brown load makes powermanagement heavily depend on software measurement and monitoring, which often suffers inaccuracy. We propose DualPower, a novel power provisioning architecture that enables green datacenters to integrate renewable power supply without grid tie inverters. To optimize DualPower operation, we propose a specially designed power management framework to coordinate workload balancing with power supply switching. We evaluate three optimization schemes (LM, PS and JO) under different datacenter operation scenarios on our trace-driven simulation platform. The experimental results show that DualPower can be as efficient as grid-tied system and has good scalability. In contrast to previous works, DualPower integrates renewable power at lower cost and maintains full availability of datacenter servers.

Data center power minimization with placement optimization of liquid-cooled servers and free air cooling

Most of today's data centers are equipped with servers that rely on air cooling, which is well known to have low cooling efficiency due to undesired air recirculation. As a result, many data centers have started to adopt liquid cooling and free air cooling for improved cooling efficiency. In this paper, we make two key observations. First, since data centers normally replace only a portion of their servers at a time, an important problem is where in the data center to place those new liquid-cooled servers for the best return on their investment. Given the complex thermal dynamics in a data center, we find, in our process of deploying liquid-cooled servers, different placement strategies lead to significantly different cooling power consumption. Second, different cooling techniques, including traditional air cooling, liquid cooling, and the emerging free air cooling, must be intelligently coordinated with dynamic workload allocation in order to minimize the cooling and server power of a data center. Based on the two observations, we propose SmartPlace, an intelligent placement algorithm that deploys liquid-cooled servers to minimize the power consumption of the data center cooling system. SmartPlace also takes into account the coordination with free air cooling and dynamic workload distribution among servers for jointly minimized cooling and server power in the entire data center. We compare SmartPlace with a state-of-the-art cooling optimization solution for two data centers with 1,280 and 10,240 servers, respectively. The results show that SmartPlace achieves up to 26.7% (with an average of 15%) less total power consumption with dynamically guaranteed application response time. Our hardware testbed results also demonstrate the effectiveness of SmartPlace. Moreover, we analyze how soon a data center can gain a full return on the capital investment of liquid-cooled servers.

Fuel Cell Powered Data Centers: In-Rack DC Generation

In this study, the National Fuel Cell Research Center is working with Microsoft to investigate a new data center concept that includes the detailed integration of fuel cell technology into server racks. The concept could eliminate the need for backup power and potentially introduce significant emissions reductions and energy savings while enhancing data center availability. The concept was experimentally verified using a hydrogen-fueled 10 kW proton exchange membrane fuel cell stack and system as the distributed direct current power source, eliminating the need for power conversion or distribution in the data center and the utility grid outside of the data center. The subsequent evaluation of a natural gas fueled 2.5 kW solid oxide fuel cells stack and system was tested and demonstrated to meet the dynamic DC power demands of servers. Both the steady state performance and the dynamic load following capability of the SOFC stack and system are evaluated and characterized.

An Auction Mechanism for Cloud Spot Markets

Dynamic forms of resource pricing have recently been introduced by cloud providers that offer Infrastructure as a Service (IaaS) capabilities in order to maximize profits and balance resource supply and demand. The design of a mechanism that efficiently prices perishable cloud resources in line with a provider’s profit maximization goal remains an open research challenge, however. In this article, we propose the Online Extended Consensus Revenue Estimate mechanism in the setting of a recurrent, multiunit and single price auction for IaaS cloud resources. The mechanism is envy-free, has a high probability of being truthful, and generates a near optimal profit for the provider. We combine the proposed auction design with a scheme for dynamically calculating reserve prices based on data center Power Usage Effectiveness (PUE) and electricity costs. Our simulation-based evaluation of the mechanism demonstrates its effectiveness under a broad variety of market conditions. In particular, we show how it improves on the classical uniform price auction, and we investigate the value of prior knowledge on the execution time of virtual machines for maximizing profit. We also developed a system prototype and conducted a small-scale experimental study with a group of 10 users that confirms the truthfulness property of the mechanism in a real test environment.

Contentionless Architecture for Optical Interconnect (CAOI) in Data Center Network

Emerging web-based and cloud computing applications are creating the need for powerful data center networks (DCN). Optical interconnects are replacing their electrical counterparts due to their benefits. An optical interconnect that eliminates contention is proposed, hereafter called CAOI. It eliminates electronic buffering and virtual output queues which reduces the latency. Fixed data rate is guaranteed under any traffic pattern. A large class of DCN applications can be supported by this architecture. Benefits of optical wavelength division multiplexing, optical code division multiple access and optical space division multiplexing are exploited. All-to-one communication is worst communication pattern encountered in any DCN application. The performance of the proposed architecture is compared for many-to-many and all-to-one communication pattern.

Data Center Power Management-Server Consolidations Techniques

Most of the companies in Information Technology and non IT companies are migrating their applications to cloud environment. Cloud is a term which is providing the most of data center modules as services to the customer to shrink the Capital Expenses. But will it really reduce the power in cloud service provider's Datacenter? This paper explains about the resourceful power management by doing server consolidations assessment based on the applications reliance and precarious before moving into virtualization and cloud which will adequate power in cloud service provider data center.

NIPD: Non-Intrusive Power Disaggregation in Legacy Datacenters

Fine-grained power monitoring, which refers to power monitoring at the server level, is critical to the efficient operation and energy saving of datacenters. Fined-grained power monitoring, however, is extremely challenging in legacy datacenters that host server systems not equipped with power monitoring sensors. Installing power monitoring hardware at the server level not only incurs high costs but also complicates the maintenance of high-density server clusters and enclosures. In this paper, we present a zero-cost, purely software-based solution to this challenging problem. We use a novel technique of non-intrusive power disaggregation (NIPD) that establishes power mapping functions (PMFs) between the states of servers and their power consumption, and infer the power consumption of each server with the aggregated power of the entire datacenter. The PMFs that we have developed can support both linear and nonlinear power models via the state feature transformation. To reduce the training overhead, we further develop adaptive PMFs update strategies and ensure that the training data and state features are appropriately selected. We implement and evaluate NIPD over a real-world datacenter with $326$ nodes. The results show that our solution can provide high precision power estimation at both rack level and server level. In specific, with PMFs including only two nonlinear terms, our power estimation i) at rack level has mean relative error of $2.18$ percent, and ii) at server level has mean relative errors of $9.61$ and $7.53$ percent corresponding to the idle and peak power, respectively.

Direct-to-chip liquid cooling for reducing power consumption in a subarctic supercomputer centre

Reduction of data centre power consumption is a timely challenge. Waste heat reuse is another focus area when developing energy efficient and sustainable data centres. And these two issues are interconnected through liquid cooling of server racks and/or direct-to-chip liquid cooling. Both of these solutions make it possible to transfer a significant proportion of the waste heat energy back to profitable use. Nevertheless, the heat reusing opportunity is not the only benefit direct-to-chip liquid cooling may offer. Another benefit is the notable reduction of power consumption related to cooling fans associated with server blades and rack-level cooling systems. To evaluate this benefit, we performed power consumption and performance measurements in a subarctic supercomputer centre hosting a cluster of 632 blade nodes. Our study concentrated on a 47-node subset that we analysed when the servers were executing the LINPACK benchmark. Our conclusion is that direct-to-chip liquid cooling can reduce the total power consumption, in this case, up to 14.4% depending on the inlet air temperature.

Power and Fault Aware Reliable Resource Allocation for Cloud Infrastructure

Cloud computing is now trending and more popular in these days for the computation and adopted by many companies like google, amazon, Microsoft etc. As the cloud size increases with increase in number of data center power consumption over a data center increases. As number of request over the data center increase with increase in load and power consumption of the data center. So the requests need to be balanced in such manner which having more effective strategy for resources utilization, request failure and improved power consumption. Cloud computing made it more complicated with respective to requests type that may increase or decrease power consumption. A recent survey on cloud computation shows that the power consumption of a server, increasing in a linear way due to utilization of resource (processors) resulting in request failure at datacenters. Request balancing in such manner without having knowledge of load over server maximize resource utilization but also increasing power consumption at server. So to overcome these issues in cloud Infrastructure as a service (IaaS), we have proposing a fault and power aware scheduling algorithm to minimize the power consumption, request failure and cost over a data center. Proposed algorithm has proven to have better performance in term of load and power efficiency as compared to previously proposed load balancing algorithm for cloud IaaS.

Power & Load Aware Resource Allocation Policy for Hybrid Cloud

Cloud computing is now trending and more popular in these days for the computation and adopted by many companies like google, amazon, Microsoft etc., As the cloud size increases with increase in number of data center power consumption over a data center increases. As number of request over the data center increase with increase in load and power consumption of the data center. So the requests need to be balanced in such manner which having more effective strategy for utilization of resources, balance load over data center and reduction in power consumption. Hybrid cloud computing made it more complicated with respective to requests type that may increase or decrease power consumption. A recent survey on cloud computation shows that the power consumption of a server, increasing in a linear way due to utilization of resource (processors). Request balancing in such manner without having knowledge of load over data center maximize resource utilization but also increasing power consumption at server. So to overcome these issues in cloud Infrastructure as a service (IaaS), we have proposing a load balancing algorithm to minimize the power consumption and cost over a data center. Proposed algorithm has proven to have better performance in term of load and power efficiency as compared to previously proposed load balancing algorithm for cloud IaaS.

Adaptive Power Efficiency Control by Computer Power Consumption Prediction Using Performance Counters

As the number of data centers has grown in pace with the burgeoning expansion in information technology (IT), reducing data center power consumption has become an important social challenge. Since IT equipment is a primary contributor to the data center power consumption, it is also an area where power conservation efforts are indispensable. Nevertheless, although the efficiency of the power supply modules integrated into computers has recently seen significant improvements, their overall efficiency primarily depends on load rates. This is particularly true under low-power load conditions, where efficiency is known to decrease drastically. Recently, power-saving techniques that work by controlling the power module configuration under low-power load conditions have been considered, and using such techniques, it is expected that further efficiency improvements can be achieved by interlocking the computer load status with the power supply configuration control in real time. In this paper, the performance counters built into computer processors are used to predict power load variations, and an equation that facilitates the prediction of server power consumption levels is defined. In a server application experiment utilizing prototype computer hardware and multiple regression analysis, we confirmed that the power prediction equation could precisely predict computer power requirements and achieve adaptive efficiency control. In order to further validate the efficiency control effect, a circuit simulation was conducted. The simulation results showed that a power loss reduction of up to 28.3% could be achieved using the model equations to control the power supply module efficiency.

Data Center Energy Consumption Modeling: A Survey

Data centers are critical, energy-hungry infrastructures that run large-scale Internet-based services. Energy consumption models are pivotal in designing and optimizing energy-efficient operations to curb excessive energy consumption in data centers. In this paper, we survey the state-of-the-art techniques used for energy consumption modeling and prediction for data centers and their components. We conduct an in-depth study of the existing literature on data center power modeling, covering more than 200 models. We organize these models in a hierarchical structure with two main branches focusing on hardware-centric and software-centric power models. Under hardware-centric approaches we start from the digital circuit level and move on to describe higher-level energy consumption models at the hardware component level, server level, data center level, and finally systems of systems level. Under the software-centric approaches we investigate power models developed for operating systems, virtual machines and software applications. This systematic approach allows us to identify multiple issues prevalent in power modeling of different levels of data center systems, including: i) few modeling efforts targeted at power consumption of the entire data center ii) many state-of-the-art power models are based on a few CPU or server metrics, and iii) the effectiveness and accuracy of these power models remain open questions. Based on these observations, we conclude the survey by describing key challenges for future research on constructing effective and accurate data center power models.

RE-UPS: an adaptive distributed energy storage system for dynamically managing solar energy in green datacenters

Datacenters, the essential infrastructures for supercomputing and cloud computing, are facing increasing pressure of capping tremendous power consumption and carbon emission. Many studies have proposed to leverage energy storage devices to shave peak power or smooth intermittent power for datacenters, respectively. However, a joint energy management of peak shaving and renewable energy harvesting in datacenters is still lacking. In this paper, we propose a new power management scheme named RE-UPS, which explores the opportunity to shave datacenter peak power demand with renewable energy. RE-UPS is based on the emerging distributed energy storage architecture and existing UPS infrastructure of datacenter. It further leverages a dynamic heuristic algorithm to determine the appropriate energy storage allocation and server power sources. The proposed energy management policies can greatly optimize the design among maximizing renewable energy harvest, shaving peak power, and maintaining UPS energy availability. Compared to the baseline power management schemes, RE-UPS can averagely improve backup energy capacity by 28 %, extend battery lifetime by 42 %, increase green energy utilization by 78 %, and reduce workload performance degradation by 13 %.

Numerical investigation and parametric study for thermal and energy management enhancements in data centers' buildings

A CFD numerical investigation of air flow and temperature distribution in data centers is presented for different operating and geometric conditions. Numerical experiments were conducted to parametrically study the effects of perforated tiles opening ratio, data center power density and locations of Computer Room Air Conditioning (CRAC) on the data center performance. Temperature distribution especially hot spot points, air flow characteristics particularly air recirculation and bypass and thermal management in data centers are evaluated in terms of the measureable overall performance parameters: supply/return heat indices (SHI/RHI), return temperature indices (RTI) and return cooling indices (RCI). Techniques of using cold aisle containments for enhancing the cooling effectiveness of data centers are investigated and evaluated. The results showed that (i) increasing the perforation ratio of perforated tails decreases the recirculation and bypass of air flow around the rack, (ii) increasing data centers' power density and supplied air flow rates increases the air bypass and air re-circulation, (iii) changing the spacing between the CRAC units and cold aisle slightly changes of air recirculation and bypass around the racks, and (iv) using appropriate cold aisle containment improves the performance of cooling of data centers, especially at high power density.

Experimental investigations of thermal managements solutions in data centers buildings for different arrangements of cold aisles containments

Cold aisles containments are used in data centers buildings to improve the thermal managements of the IT servers. In the present study, an experimental investigation of air flow and thermal management of data center buildings is conducted for different arrangements of cold aisles containments. A scaled physical model simulating real data centers is designed using scale modeling theory and constructed for the sake of the present experimental work. The data center scaled model accommodate a row of three racks; each rack included four servers for testing purposes. Three arrangements of cold aisles are tested: (i) free open cold aisle, (ii) semi-enclosed cold aisles where the aisles are enclosed from the sides, and (iii) full enclosed cold aisles, where the aisles are enclosed from sides and top. Front and rear racks temperatures profile, servers temperatures and performance measuring parameters Supply/Return Heat Index (SHI/RHI) are used to evaluate the thermal management/performance of the data centers racks. Experiments were conducted at different racks power densities. The results showed that (i) increasing data centers power density and supplied air flow rates improves SHI but increases servers temperature, (ii) using semi-enclosed cold aisles relatively improve the thermal performance of the data centers, (iii) using full enclosed cold aisles dramatically improve the thermal performance of the data center servers, and (vi) the percentage of enhancement in data centers performance due to using cold aisle containments dramatically increases with increasing the power density.

Holistic multiobjective planning of datacenters powered by renewable energy

Energy efficiency is a major concern to datacenter operators, because the large amounts of energy used by parallel computing infrastructures increases costs and affects the electricity grid. Datacenter power consumption can be reduced by applying intelligent control techniques to dynamically adjust power demand, but this is hampered by conflicting objectives. For instance, the workload can be controlled to adjust power, but at the expense of service quality. Or, the cooling infrastructure demand can be manipulated without affecting workloads, but at the risk of shifting the datacenter temperature outside the acceptable limits. This paper proposes a multiobjective, evolutionary approach to solving the problem of energy-aware task scheduling in datacenters. Our approach takes into account three problem objectives (power consumption, temperature, and quality of service) when both computing and cooling infrastructures are holistically controlled. We report the two best solutions to each of these problem objectives, as well as the selected trade-off solutions between them.

HP plans to power data centers with renewable energy

HP plans to power data centers with renewable energy

Virtual Machine Load Balancing: State-of-the-Art and its Research Challenges

Cloud computing is the most demanding technologies in today’s era due to the reduction of cost and complexity of applications, as a result number of users demanding more services of cloud. Thus, for overall cloud performance load balancing in the cloud is an important research area. Cloud handles large number of users as it is working on powerful datacenters. So for better efficiency it has to be featured with reliable load balancer for handling the load in the cloud. Many load balancing algorithms have been proposed for the better management of user request among cloud resources. This paper presents our research with a survey of various existing load balancing algorithms along with state of the art and its research challenges.

HEB

Today, an increasing number of applications and services are being hosted by large-scale data centers. The massive and irregular load surges challenge data center power infrastructures. As a result, power mismatching between supply and demand has emerged as a crucial issue in modern data centers which are either under-provisioned or powered by intermittent power sources. Recent proposals have employed energy storage devices such as the uninterruptible power supply (UPS) systems to address this issue. However, current approaches lack the capacity of efficiently handling the irregular and unpredictable power mismatches. In this paper, we propose Hybrid Energy Buffering (HEB), the first heterogeneous and adaptive strategy that incorporates super-capacitors (SCs) into existing data centers to dynamically deal with power mismatches. Our techniques exploit diverse energy absorbing characteristics and intelligent load assignment policies to provide efficiency- and scenario- aware power mismatch management. More attractively, our management schemes make the costly energy storage devices more affordable and economical for datacenter-scale usage. We evaluate the HEB design with a real system prototype. Compared with a homogenous battery energy buffering system, HEB could improve energy efficiency by 39.7%, extend UPS lifetime by 4.7X, reduce system downtime by 41% and improve renewable energy utilization by 81.2%. Our TCO analysis shows that HEB manifests high ROI and is able to gain more than 1.9X peak shaving benefit during an 8-years period. It allows datacenters to adapt to various power supply anomalies, thereby improving operational efficiency, resiliency and economy.