Vikneshan Sundaralingam

Experimental characterization of cold aisle containment for data centers

The data center industry has experienced significant growth over the last decade, mainly due to the increased use of the internet for our day to day activities such as e-commerce, social media, video streaming and healthcare. This growth in demand results in higher energy costs, as data centers can be energy intensive facilities. A significant portion of the energy used in data centers is for cooling purposes. Hence, it is one of the important areas of optimization to be addressed to create more efficient data centers. Amongst the many ways to increase data center efficiencies, airflow management is a key solution to many existing data centers. Fundamentally, there are three main schemes: hot-aisle containment, cold-aisle containment and exhaust chimney containment. This papers focus is to experimentally characterize the following cold aisle configurations: open aisle, partially contained aisle and fully contained aisles. Experimental data presented to evaluate the effectiveness of the different configurations are rack inlet contour plots, tile and rack flow rates, pressure measurements, and server CPU temperatures.

Effect of rack server population on temperatures in data centers

To reduce energy wastage during equipment upgrades that require changing the number of servers inside a rack, this paper investigates the effect of the server population of a rack on air temperatures in a data center facility. A rack-level containment system was implemented and air temperatures in cold and hot aisles with different numbers of servers were measured. A computational fluid dynamics (CFD)-based numerical model is used to study the air flow field. In addition, this paper includes the investigation of the effect of void locations on CPU temperatures. The study reveals that the server population inside a rack has a significant impact on air temperatures.

Dynamics of cold aisle air distribution in a raised floor data center

Server manufacturers test compute equipment under ideal conditions such as quiescent surroundings, with no adverse pressure gradients at the inlet, which, hardly exist in a real data center. The dynamics of cold air delivery to the server can significantly impact server cooling and performance, especially at higher rack densities. Presently, very little is known on the physics of cold air distribution in a rack in a high density data center environment. The experimental work reported in this paper provides an insight into the raised floor cold air delivery system currently adopted for rack cooling in a hot aisle/cold aisle data center. Air flow measurements using Particle Image Velocimetry (PIV) are performed in the cold aisle at the inlet of a single rack. Two cases of perforated tile air delivery rates are investigated. The first case with no discharge from the tile represents the ideal case under which the servers are typically tested by manufacturers. The second case represents a realistic high density data center environment, wherein 0.755 m3/s (∼1600 CFM) of air is discharged through the perforated tile. The analysis of the flow dynamics at the perforated tile surface, rack inlet and cold aisle boundaries is reported for both cases. A distinct change in the cold aisle air velocities and flow pattern at the interrogation boundaries is observed for both the cases.

Effect of server load variation on rack air flow distribution in a raised floor data center

This paper presents experimental studies on server level rack air distribution for a preset perforated tile air flow rate. A series of experiments are performed using a 22.8 kW server simulator placed in a raised floor data center facility. The rack air flow rate is varied by adjusting the fan speed of the server simulator. Particle Image Velocimetry (PIV) technique is used to capture the air flow pattern at various locations in the cold aisle at the inlet of the server simulator. The PIV images are recorded at various locations and later combined to get the complete air distribution map across the rack inlet. Various cases of rack air flow distributions are investigated by varying the server simulator fan speed settings for given perforated tile air flow rate. A significant change in the air distribution pattern is observed for various cases investigated.

Server Heat Load Based CRAC Fan Controller Paired With Rear Door Heat Exchanger

This paper discusses the preliminary design of a controller for the cooling air conditioning units (CRAC) fan speed based on server loads. The CRAC is paired with a rear door heat exchanger. The main purpose in implementing this controller is the reduce energy consumption and to have finer control in the provision of cooling within a data center. Tools that will be essential for developing this controller are CRACs with variable frequency drives (VFD), BCMS power monitoring, data archiving software (PI from OSISOFT) and MATLAB. For this experiment, the zone of interest in the data center consists of 10 racks of IBM Blade Centers with a maximum electrical load of approximately 24 kW each and a total of 3360 nodes. Results are compared with the conventional cooling method to quantify energy savings and corresponding chiller coefficient of performance (COP).Copyright

Airflow Management in a Contained Cold Aisle Using Active Fan Tiles for Energy Efficient Data-Center Operation

Generally, passive perforated tiles are used in a data center and the supplied airflow rate is underprovisioned; thus, the balance of the server air requirement is met by the hot air in the room, resulting in higher server inlet temperatures. Full provisioning of the supplied airflow rate and containing the cold aisle is expected to minimize the hot air leakage in the cold aisle, resulting in uniform and lower server inlet temperatures. Thus, the supply air temperature can be raised, resulting in energy savings at the chiller plant. Supplying extra air can be achieved using active perforated tiles, having multiple fans installed on them. In this paper, the underprovisioned case using passive tiles and the fully provisioned case using active tiles are investigated for both open and contained aisle conditions. Thermal field measurements suggest lower and uniform server inlet temperatures for the case with contained aisle as compared to open aisle and for the fully provisioned case using active tiles as compared to the underprovisioned case using passive tiles. System-level energy calculations suggests that containing the cold aisle results in lower (improved) power usage effectiveness (PUE); however, use of active tiles does not seem to improve the PUE.

Investigation of exergy destruction in CFD modeling for a legacy air-cooled data center

To keep up with the growing energy demand, legacy aircooled data centers have begun to implement simple but effective energy efficiency strategies or best practices. Improving “air management”, optimizing the delivery of cool air and the collection of waste heat, are among these efficiency strategies. One of the most powerful tools for identifying energy wasteful practices is the application of the second law of thermodynamics to estimate the exergy destruction in a system. Exergy is defined as the “available energy” in a system, hence wasteful practices are said to “destroy exergy.” Systematic analysis of data centers using this approach will identify the major contributors to wasteful energy practices including premature mixing of hot and cold air streams. In this study, a numerical model was developed for estimation of the data center airspace exergy destruction following the three-dimensional CFD simulation of room airspace. A second exergy destruction model available in the literature was implemented as well. Both models were tested on one simplified case with available analytical solution. They were then utilized to estimate the airspace exergy destruction in an existing research data center. Both approaches proved to be adequate, although the one proposed in this work presented significant sensitivity to numerical inaccuracies compared to the second model, even leading to zones of artificial negative exergy destruction.

Experimental Characterization of Various Cold Aisle Containment Configurations for Data Centers

The data center industry has experienced significant growth over the last decade, mainly due to the increased use of the internet for our day to day activities such as e-commerce, social media, video streaming, and healthcare. This growth in demand results in higher energy costs, as data centers can be energy intensive facilities. A significant portion of the energy used in data centers is for cooling purposes. Hence, it is one of the important areas of optimization to be addressed to create more efficient data centers. Among the many ways to increase data center efficiencies, air flow management is a key solution to many existing data centers. Fundamentally, there are three main schemes: hot-aisle containment, cold-aisle containment, and exhaust chimney containment. This papers focus is to experimentally characterize the following cold aisle configurations: open aisle, partially contained aisle, and fully contained aisles. Experimental data presented to evaluate the effectiveness of the different configurations are rack inlet contour plots, tile and rack flow rates, pressure measurements, and server central processing unit (CPU) temperatures.

Effect of Supply Air Temperature on Rack Cooling in a High Density Raised Floor Data Center Facility

In this paper we experimentally investigate the effect of supply air temperature on rack cooling in a high density raised floor data center facility. A series of experiments are performed on a 42 U (1-U = 4.45 cm) rack populated with 1-U servers. Desired rack heat loads are achieved by managing the distribution of server compute load within the rack. During the present experiments, temperatures at various locations in the hot and cold aisle corresponding to the rack air inlet and outlet are recorded. The temperatures are measured using a grid consisting of 256 thermocouples. The temperature measurements are further complimented with the flow field at the rack inlet. Particle Image Velocimetry (PIV) technique is used to capture the flow field at the rack inlet. The temperature maps in concert with the PIV flow field help in quantifying the rack cooling effectiveness. The temperature and flow measurements are measured for various cases by altering the supply air temperatures and perforated tile flow rates. The results are analyzed and compared with the ASHRARE recommended guidelines to arrive at the optimum supply air temperature. A perceptible change in the temperature and flow distribution is observed for the six cases investigated.© 2011 ASME

Modeling Thermal Mass of a Data Center Validated With Actual Data due to Chiller Failure

The task of minimizing the downtime of a data center is becoming increasingly important due to the necessity of availability and maintaining the integrity of the data being handled by the data center. Consequently, a model used to predict the thermal response of a data center would be useful information in designing mechanisms to minimize the downtime during a failure or to serve as an alternative analysis method other than CFD. This paper will focus on a thermodynamic approach of predicting the thermal response of the data center space with the use of lumped system analysis. The model will be developed and validated using actual data from a chiller failure event in the CEETHERM Data Center Laboratory. Events in sequence are: (i) Chiller failure, (ii) Data center shutdown due to critical temperatures and (iii) Chiller restored. To illustrate, the data center section of interest consists of 10 racks of servers (maximum capacity of 24kW for each rack) with a total of 3360 nodes and is chilled using chilled water from the building chiller, through which the cooling resources are distributed using a rear door heat exchanger and a cooling room air conditioning unit (CRAC). The relevant and important data that was recorded in this failure are the: (1) Server inlet temperatures, (2) CPU temperatures, (3) CRAC supply and return air temperatures, (4) Chiller supply and return water temperature, (5) Chiller flow rate, (6) Data center space temperature and humidity, (7) Server power draw and (8) CRAC fan speeds.Copyright