websights

Fill out the form to download

Required field
Required field
Not a valid email address
Required field
Required field

  • Set up your own cloud-native simulation in minutes.

    • Documentation

    SimScale DocumentationValidation CasesValidation Case: Airflow in a Data Center

    Validation Case: Airflow in a Data Center

    The airflow in a data center validation case belongs to fluid dynamics. This test case aims to validate the following:

    • Velocity profile

    The validation is performed against experimental data and simulations presented by Wibron, Ljung, and Lundström\(^{1}\), who studied the flow field behavior within a data center module.

    View Project

    Geometry

    The geometry consists of a data center module with four Computer Room Air Conditioning (CRAC) units, as well as ten server racks, as shown in the schematics below:

    schematics data center module validation
    Figure 1: Schematics of the data center module, identifying server racks (R) and CRAC units (C). Adapted from [Wibron, Ljung, and Lundström]\(^{1}\).

    Furthermore, the authors also provide an isometric image of the data center module, showing its dimensions.

    data center crac units validation dimensions
    Figure 2: Dimensions of the data center module. Adapted from [Wibron, Ljung, and Lundström]\(^{1}\)

    The schematics provided in the study contain the dimensions of the module, however they do not provide any additional details about the dimensions of the equipment.

    Therefore, to obtain the final geometry used in this validation case, the isometric image shown in Figure 2 was manually measured to determine the missing dimensions. Figure 3 shows the resulting geometry in SimScale:

    data center validation geometry
    Figure 3: By manually measuring the schematics from the paper, we can closely estimate the remaining dimensions.

    Analysis Type and Mesh

    Tool Type: OpenFOAM®

    Analysis Type: Steady-state, incompressible Convective heat transfer analysis

    Turbulence Model: k-omega SST.

    Mesh and Element Types: The meshes used in this validation case were created in SimScale with the hex-dominant (parametric) algorithm. To ensure that the results are mesh-independent, a total of three successively finer meshes were used to run the simulations. The results were found to be mesh-independent.

    Table 1 presents relevant metrics and details of the resulting meshes:

    DomainIdentificationMesh TypeCellsElement Type
    Data center moduleCoarse meshHex-dominant (parametric)1273143D hexahedral
    Data center moduleModerate meshHex-dominant (parametric)8205613D hexahedral
    Data center moduleFine meshHex-dominant (parametric)60595783D hexahedral
    Table 1: For mesh sensitivity studies, it is important to evenly refine the domain, obtaining finer meshes

    Find below, in Figure 4, the fine hex-dominant (parametric) mesh generated in SimScale.

    Figure 4: Fine hexahedral mesh. A region refinement is applied on the hot side of the server racks to capture the sharp velocity and temperature gradients.

    Simulation Setup

    Material:

    The fluid material for this simulation is air. In terms of material properties, the authors did not indicate any values, except for the specific heat. Therefore, the following material settings are used in this validation case:

    • Viscosity model: Newtonian
    • \((\nu)\) Kinematic viscosity: 1.5295e-5 \(m^2/s\)
    • \((\rho)\) Density: 1.1965 \(kg/m^3\)
    • Thermal expansion coefficient: 0.00343 \(\frac {1} {K}\)
    • \(T_0\) Reference temperature: 273.15 \(K\)
    • \((Pr_{lam})\) Laminar Prandtl number: 0.713
    • \((Pr_{t})\) Turb. Prandtl number: 0.85
    • Specific heat: 1004.4 \(\frac {J} {kg.K}\)

    Boundary Conditions:

    In the reference study, the authors provided the volumetric flow rate and the temperature at the inlet of the CRAC units. Furthermore, the volumetric flow rate and heat dissipation are also given for each server rack.

    Following the same approach used in the simulations from the reference paper, each server rack has an outlet (cold side) and an inlet (hot side). The temperature at the server inlets is calculated based on the flow rate through the server rack, the temperature at the cold side, and the heat dissipation. The formula is described in this article.

    In the table below, the configuration for both velocity and pressure are given at each of the boundaries:

    BoundaryBoundary TypeBoundary Condition Description
    Crack unit 1 – InletVelocity inlet0.61 \(m/s\) at 19.7 \(ºC\)
    Crack unit 2 – InletVelocity inlet0.62 \(m/s\) at 20 \(ºC\)
    Crack unit 3 – InletVelocity inlet0.59 \(m/s\) at 19.7 \(ºC\)
    Crack unit 4 – InletVelocity inlet0.62 \(m/s\) at 19.9 \(ºC\)
    Outlets – CRAC unitsPressure outletFixed at 0 gauge \(Pa\)
    Rack 1 – Hot sideCustom0.4 \(m/s\) inlet at 35.171 \(ºC\)
    Rack 2 – Hot sideCustom0.46 \(m/s\) inlet at 33.228 \(ºC\)
    Rack 3 – Hot sideCustom0.47 \(m/s\) inlet at 32.864 \(ºC\)
    Rack 4 – Hot sideCustom0.45 \(m/s\) inlet at 33.412 \(ºC\)
    Rack 5 – Hot sideCustom0.26 \(m/s\) inlet at 20.000 \(ºC\)
    Rack 6 – Hot sideCustom0.41 \(m/s\) inlet at 34.255 \(ºC\)
    Rack 7 – Hot sideCustom0.38 \(m/s\) inlet at 36.013 \(ºC\)
    Rack 8 – Hot sideCustom0.43 \(m/s\) inlet at 34.347 \(ºC\)
    Rack 9 – Hot sideCustom0.42 \(m/s\) inlet at 34.856 \(ºC\)
    Rack 10 – Hot sideCustom0.39 \(m/s\) inlet at 35.614 \(ºC\)
    Racks 1 through 10 – Cold sideVelocity outletFace normal value for velocity, equal to the hot side of each rack
    Internal wallsWallNo-slip condition, adiabatic
    External wallsWallNo-slip condition, adiabatic
    Table 2: Overview of the boundary conditions. Wall functions were used for the wall treatment.

    Result Comparison

    In the reference study\(^{1}\), the air velocity was measured over five lines (L1 through L5). For all lines, measurements were made 0.5, 1, and 1.5 \(m\) above the ground. For L4 and L5, one additional measurement was taken at 2 meters height.

    Since no precise locations were given for the probe lines, the figure below was used to estimate their coordinates.

    schematics data center experimental results
    Figure 5: Schematics showing where the experimental results are evaluated. L1 through L3 are on the hot side of the racks.

    Velocity Profiles

    Find in the images below a comparison between the velocities obtained in SimScale, the experimental data, and the simulation results from the reference paper\(^{1}\).

    results data center validation velocity magnitude
    Figure 6: Velocity plots showing the results from the reference study and SimScale results. Lines L1 through L3 are on the hot side of the server racks.

    The fine mesh results from SimScale show the same trends as the experimental and simulation results from the reference study. Given the lack of a precise description of the data center module and air properties, differences in the results are expected.

    Moreover, the fine mesh used in SimScale has 6059578 cells, whereas the fine mesh from the paper has only 1486077 cells. Figure 7 contains a similar comparison, now for lines L4 and L5.

    result comparison data center module cfd validation
    Figure 7: Result comparison SimScale and the reference study. Lines L4 and L5 are located on the cold side of the server racks.

    For lines L4 and L5, the velocity profiles obtained with SimScale are also showing the same trends as the results from the reference study. This is expected, based on the assumptions that were made during the case setup.

    data center module velocity field and vectors validation
    Figure 8: Cutting plane in the middle of the room showing the velocity field.

    The flow is slightly unsymmetrical, due to differences in the flow rates through the CRAC units and server racks.

    References

    Last updated: March 15th, 2021

    Contents