Evaluating CFD Results of a Data Center

evaluating_cfd_results_of_a_data_center.jpg

Recirculation of air at the end of a data center aisle, as visualized in Simulation CFD.

In this section, the performance of a data center will be evaluated by visualizing the flow and thermal distribution of air entering equipment in the room. Extracting data from those views will establish a metric to quantify performance and lead to further interrogation that reveals opportunities for improvement.

The temperature of air entering computer equipment will impact its performance and reliability.  Delivering a consistent amount of cool air is very important in data center design.  Accounting for the path of air and its temperature as it moves through a data center enables designers to predict and optimize equipment placement. 

A typical data center room is comprised of computer equipment in rows of racks (servers in cabinets) along with an air delivery and conditioning system .

(1) Cool air entering from floor tile  (2) Equipment Racks  (3) Exhaust air exiting from ceiling

(1) Cool air entering from floor tile

(2) Equipment Racks

(3) Exhaust air exiting from ceiling

The energy used to move and condition air is wasted when the cooler air does not move directly through equipment.  The recirculation of air around the room or equipment will:

  • Increase equipment air intake temperatures.
  • Reduce the air delivery and conditioning systems efficiency.
  • Degrade equipment performance and lifespan.

Visualizing Global Temperatures

The temperature of air entering equipment is a critical value that can be viewed in Simulation CFD with Global results.  Global results are displayed on model surfaces and will be visible on equipment intake and exhausts.

Global temperature results are plotted on fan devices that represent rack equipment; cool aisles (1) hot aisles (2).

Global temperature results are plotted on fan devices that represent rack equipment; cool aisles (1) hot aisles (2).

Cool air intake of a single row of a datacenter.

Cool air intake of a single row of a datacenter.

 

 

Looking for Unwanted Temperature Gradients

Inconsistent air intake temperatures produce variations in equipment performance and lifespan, and should be avoided.  Understanding the air intake temperature variation and its cause can be accomplished by interpreting the results visualizations and effectively using the data extraction tools of Simulation CFD.

Aligning a results Plane to the equipment intake provides a visualization of the temperature gradients (similar to the global view). Additionally, XY Plots, the Bulk Calculator, and Traces can quantify the variation in temperature and help identify its cause.

 

Quantifying Temperature Profiles with XY Plots

Extracting the temperature gradient with an XY Plot reveals its magnitude by plotting values directly to a graph.


Front view of server rack intake temperatures with a dotted line where data is extracted.

The plot depicts a non-uniform temperature distribution between the center and outermost racks.  Red arrows are used to indicate the temperature difference found at the outermost racks compared to the center racks.

 

Quantifying Temperature Differences with Bulk Results

The average air temperature entering each rack (as opposed to the points along a line of the XY plot) can be determined with Bulk results.

Bulk results are used here to determine the average temperature entering each rack.  All solids are hidden here to isolate the fluid region of each rack.

Bulk results are used here to determine the average temperature entering each rack.  All solids are hidden here to isolate the fluid region of each rack.

Bulk data can be viewed in its respective dialog or saved and manipulated to provide a report and chart depicting the average temperature difference between racks.

Position Average Intake Temperature (F)
1 88
2 70
3 63
4 63
5 63
6 63
7 67
8 98

Bulk results are used to populate a table and chart with the average air temperatures entering each rack.

 

At this point the performance difference between each rack has been quantified and can be used as a metric to track design modifications.  The data itself, however, does not provide insight on the cause of the intake temperature variation.  This will be accomplished by tracing the path taken by air from the hottest regions.

Using Traces to Understand Root Causes

The best strategy is to start the Traces at the hottest locations of the server intakes to visualize the path that the hot air has taken.

TIP: Traces do not have to be placed at inlets and outlets, they can be added to any Plane and the software will determine the upstream and downstream path of the trace from that location.

Trace particle points (circled) are added to the hottest region, “seeding” their path.

Trace particle points (circled) are added to the hottest region, “seeding” their path.

 

The path of air moving through the trace particle points (circled above).

The path of air moving through the trace particle points (circled above).

 

The path taken by the air entering the hottest regions of the equipment recirculates from the exhaust back to the intake, increasing air temperatures at each pass of the recirculation loop.  Eliminating this condition with hot and cold aisle containment walls would reduce the outermost equipment rack intake temperatures, producing improved and more consistent equipment performance and life spans.

Recirculation can not only occur between the equipment exhaust and intake;  it can also occur within the room.

Air enters the space (1), moves past server intakes (2) and away from the equipment row (3).  The air recirculates in the room (4), increasing in temperature before returning to the equipment row (5).

Air enters the space (1), moves past server intakes (2) and away from the equipment row (3).  The air recirculates in the room (4), increasing in temperature before returning to the equipment row (5).

 

Conclusion

Simulation CFD provides designers with insight on air flow paths and temperature gradients that guide the optimization of data center performance.  Results visualization and data extraction tools quantify operating characteristics that can be compared with various design concepts (e.g., hot aisle and cold aisle containment strategies) to converge on an economical solution which conserves energy and eliminates recirculation.