Webinar Data Center Cooling
The recent boom in the computer industry has led to a rise of the need of large rooms in order to store, power and operate computer servers and their associated components. Primarily, a server room is a room used for executing enterprise applications that require massive computing resources at run-time, such as banking software, search engines or social networking applications.
Server rooms are stacked up in what is called a data center. Having applications that provide communication and services round the clock, a data center is designed to operate at a continuous supply of electrical power. This causes the servers to heat up and disrupts their proper functioning. As a result, adequate cooling infrastructure is required to make sure that the servers give maximum output with minimum energy consumption.
The current project entails the simulation of air temperature and velocity inside a server room using a thermo-fluid analysis type on SimScale. Each server panel is considered to be a source of heat, maintained at a constant temperature.
A comparative study may be performed by using different configurations of the inlet and outlet. For the purpose of this project, we consider only one case to get accustomed to the simulation process.
The CAD model of the server room has been uploaded to the SimScale platform in .STEP format. A regular cuboidal room is considered along with an array of racks.
The supply vent (inlet) is located at ground level on the side wall, whereas the exhaust (outlet) is placed close the to ceiling at the opposite end of the room.
Exterior view of server room
Interior view of server room with rack array
The imported geometry is meshed using the highly automated hex-dominant meshing operation for internal fluid flow. Due to the large geometry, the mesh size is set to coarse and 4 cores are used for the computation. Using the feature brusselTopological Entity Sets, we grouped various elements of the geometry into Inlet, Outlet, Rack and Walls.
Overall meshed geometry
Meshing mapped onto rack array
Subsequent to meshing, the simulation is set up using the analysis type - Natural Convective Heat Transfer.
To draw a comparison, two different simulations have been set up; the first makes a rough estimation by assuming a laminar flow field. On the other hand, the second one using a k-epsilon RANS turbulence model.
Moreover, the two cases operate under different boundary conditions. The first simulation has the room walls defined as adiabatic, whereas the second involves a fixed temperature assigned to the walls. This clearly demonstrates how SimScale can help answer ‘What if’ scenarios in a quick and constructive manner.
The simulation is executed using 16 cores and takes approximately 20 minutes.
Results and Conclusions
Velocity streamlines of airflow
As can be seen from the scale shown in the image above, we can analyze the velocities of airstreams flowing around the racks and across the room.
Vector glyphs for airflow
The flow is represented by vector glyphs in the above image. As a measure to overcome the challenge of visualizing the flow on boundary surfaces using CFD, a glyph filter is used.The length of the glyph is directly proportional to the velocity of the flow stream and its orientation depicts the direction of flow.
Temperature contour coupled with velocity streamlines
The simulation results help compute the required power of the cooling system under different operating conditions. Also, they provide us with a good insight into the resultant velocity of airflow and temperature field in the server room. Ergo, various room layouts with different vent-exhaust configurations can be evaluated during an early design phase, saving the time, effort and resources spent on physical testing.