Rongliang Zhou

Towards the design and operation of net-zero energy data centers

Reduction of resource consumption in data centers is becoming a growing concern for data center designers, operators and users. Accordingly, interest in the use of renewable energy to provide some portion of a data centers overall energy usage is also growing. One key concern is that the amount of renewable energy necessary to satisfy a typical data centers power consumption can lead to prohibitively high capital costs for the power generation and delivery infrastructure, particularly if on-site renewables are used. In this paper, we introduce a method to operate a data center with renewable energy that minimizes dependence on grid power while minimizing capital cost. We achieve this by integrating data center demand with the availability of resource supplies during operation. We discuss results from the deployment of our method in a production data center.

Optimization and control of cooling microgrids for data centers

Reliable operation of todays data centers requires a tremendous amount of electricity to power both the IT equipment and the supporting cooling facilities. As much as half of total data center electricity consumption can be attributed to the cooling systems required to maintain the thermal status of IT equipment. In order to lower the electricity usage of the cooling system and hence reduce the data center environmental footprint, alternative cooling resources, such as water and air-side economizers, are being exploited to supplement or replace the traditional chilled water based cooling schemes. The various cooling resource options, together with the mechanisms to distribute and deliver the cooling resource to IT equipment racks, constitute a cooling microgrid. In this paper, we present a holistic perspective for the optimization and control of the data center cooling microgrid. The holistic approach optimizes the sourcing and distribution of cooling resources from the site portfolio in response to real-time weather changes and site demand. The cooling microgrid optimization and control framework has been implemented in a research data center; we estimate that the framework cuts the yearly cooling costs by 30%.

Data center cooling management and analysis - a model based approach

As the hub of information aggregation, processing, and dissemination, todays data centers consume significant amount of energy. The data center electricity consumption mainly comes from the IT equipment and the supporting cooling facility that manages the thermal status of the IT equipment. The traditional data center cooling facility usually consists of chilled water cooled computer room air conditioning (CRAC) units and chillers that provide chilled water to the CRAC units. Electricity used to power the cooling facility could take up to a half of the total data center electricity consumption, and is a major contributor to the data center total cost of ownership. While the data center industry has established the best practice to improve the cooling efficiency, the majority of it is rule of thumbs providing only qualitative guidance. In order to provide on demand cooling and achieve improved cooling efficiency, a model based description of the data center thermal environment is indispensable. In this paper, a computationally efficient multivariable model capturing the effects of CRAC units blower speed and supply air temperature (SAT) on rack inlet temperatures is introduced, and model identification and reduction procedures are discussed. Using the model developed, data center cooling system design and analysis such as thermal zone mapping, CRAC units load balancing, and hot spot detection are investigated.

Modeling and Control for Cooling Management of Data Centers With Hot Aisle Containment

In traditional raised-floor data center design with hot aisle and cold aisle separation, the cooling efficiency suffers from recirculation resulting from the mixing of cool air from the Computer Room Air Conditioning (CRAC) units and the hot exhaust air exiting from the back of the server racks. To minimize recirculation and hence increase cooling efficiency, hot aisle containment has been employed in an increasing number of data centers. Based on the underlying heat transfer principles, we present in this paper a dynamic model for cooling management in both open and contained environment, and propose decentralized model predictive controllers (MPC) for control of the CRAC units. One approach to partition a data center into overlapping CRAC zones of influence is discussed. Within each zone, the CRAC unit blower speed and supply air temperature are adjusted by a MPC controller to regulate the rack inlet temperatures, while minimizing the cooling power consumption. The proposed decentralized cooling control approach is validated in a production data center with hot aisles contained by plastic strips. Experimental results demonstrate both its stability and ability to reject various disturbances.Copyright

Data Center Cooling Efficiency Improvement Through Localized and Optimized Cooling Resources Delivery

Data centers are large computing facilities that can house tens of thousands of computer servers, storage and networking devices. They can consume megawatts of power and, as a result, reject megawatts of heat. For more than a decade, researchers have been investigating methods to improve the efficiency by which these facilities are cooled. One of the key challenges to maintain highly efficient cooling is to provide on demand cooling resources to each server rack, which may vary with time and rack location within the larger data center. In common practice today, chilled water or refrigerant cooled computer room air conditioning (CRAC) units are used to reject the waste heat outside the data center, and they also work together with the fans in the IT equipment to circulate air within the data center for heat transport. In a raised floor data center, the cool air exiting the multiple CRAC units enters the underfloor plenum before it is distributed through the vent tiles in the cold aisles to the IT equipment. The vent tiles usually have fixed openings and are not adapted to accommodate the flow demand that can vary from cold aisle to cold aisle or rack to rack. In this configuration, CRAC units have the extra responsibilities of cooling resources distribution as well as provisioning. The CRAC unit, however, does not have the fine control granularity to adjust air delivery to individual racks since it normally affects a larger thermal zone, which consists of a multiplicity of racks arranged into rows. To better match cool air demand on a per cold aisle or rack basis, floor-mounted adaptive vent tiles (AVT) can be used to replace CRAC units for air delivery adjustment. In this arrangement, each adaptive vent tile can be remotely commanded from fully open to fully close for finer local air flow regulation. The optimal configuration for a multitude of AVTs in a data center, however, can be far from intuitive because of the air flow complexity. To unleash the full potential of the AVTs for improved air flow distribution and hence higher cooling efficiency, we propose a two-step approach that involves both steady-state and dynamic optimization to optimize the cooling resource provisioning and distribution within raised-floor air cooled data centers with rigid or partial containment. We first perform a model-based steady-state optimization to optimize whole data center air flow distribution. Within each cold aisle, all AVTs are configured to a uniform opening setting, although AVT opening may vary from cold aisle to cold aisle. We then use decentralized dynamic controllers to optimize the settings of each CRAC unit such that the IT equipment thermal requirement is satisfied with the least cooling power. This two-step optimization approach simplifies the large scale dynamic control problem, and its effectiveness in cooling efficiency improvement is demonstrated through experiments in a research data center.Copyright

Estimating data center thermal correlation indices from historical data

In order to better manage the cooling infrastructure in a data center with multiple computer room air conditioning (CRAC) units, the relationship between CRAC settings and temperature at various locations in the data center needs to be accurately and reliably determined. Usually this is done via a commissioning process which is both time consuming and disruptive. In this paper, we describe a machine learning based technique to model rack inlet temperature sensors in a data center as a function of CRAC settings. These models can then be used to automatically estimate thermal correlation indices (TCI) at any particular CRAC settings. We have implemented a prototype of our methodology in a real data center with eight CRACs and several hundred sensors. The temperature sensor models developed have high accuracy (mean RMSE error is 0.2°C). The results are validated using manual commissioning, demonstrating the effectiveness of our techniques in estimating TCI and in determining thermal zones or regions of influence of the CRACs.

Office building model identification and control design

This paper presents the identification of a lumped thermal model for one floor of a large office building consisting of partitioned cubicles and conference rooms. The space is instrumented with a large number of interior temperature sensors as well as supply-air temperature and flowrate sensors. Roof top solar radiation data is also available. The first step in the model identification process is to cluster the interior temperature measurements into zones. The zones form a thermal connectivity graph based on the physical location of the clusters. Each zone corresponds to a node in the graph with a corresponding thermal capacitance. The zones connect to adjacent zones and to the ambient via thermal resistances. Due to large windows on two walls and the ceiling, the model also needs to account for the radiant heat input into the space. We identify thermal capacitance, thermal resistance, and radiant heat transfer coefficients in the model by matching the predicted temperature trajectory with the measured data. As validation, the model shows reasonable prediction of temperatures on dates not used for identification. Using the identified model, we also consider the temperature control problem with real-world ambient temperature and solar data. By using our previously reported passivity-based temperature controller, we achieve tighter temperature regulation while consuming less energy as compared with the existing controller.