Data Processing Centres, also called Data Centers, are environmentally controlled facilities dedicated to the storage and processing of large-scale data. As critical infrastructures, their security requirements are increasing. Within the design and operational requirements, the air management part plays a key role. Today’s IT equipment reaches very high power ratings, and the proper arrangement of air intake and exhaust air equipment can determine the efficiency and safe operation of the facility.
A very common technique to achieve better air management is to confine the aisles where the air is drawn in by the racks (cold aisle) or where the equipment exhausts the air (hot aisle). This method greatly reduces the recirculation of hot air at the inlet of the racks, reducing the risk of excessive local heating that can cause problems for IT equipment.
When designing Data Centers, to optimise operating conditions or study failure cases, CFD Simulation is a very useful tool. In this post we will show with a CFD study the improvements obtained in a large Data Center by confining the hot and cold aisles, which will allow us to make a better decision in the engineering phase.
How is the CFD model of a Data Center prepared?
We start by modelling the geometry of the data centre, including the racks and the air supply and exhaust systems. To perform the CFD study, the geometry must be divided into small cells or elements, which allows the fluid mechanics equations to be solved using numerical computation. This process is known as meshing. The following image shows a detail of the geometry of the room and its corresponding meshing:
Next, we define the physics of the problem and apply the corresponding boundary conditions. Specifically, we define a cooling flow rate of 1 m3/s per rack. This is fed into the room at 20ºC from a raised floor in the cold aisles. Each rack has an assigned power of 12kW, which will heat the absorbed air and expel it to the hot aisle. The hot aisle has an extraction system at the top.
Once the model is prepared, a finite volume solver is used to solve the equations until a steady state is reached. The analysis of a transient that would allow us to see the time evolution of the variables would also be possible and would provide information of great interest, but in this example we will focus on the stationary maps.
CFD simulation results
CFD study of the unconfined Data Center
The results obtained without confining any of the corridors are shown in the following pictures. They represent a cross section, first the flow velocity and below the temperature:
The temperature map shows some areas that are warmer than desirable, especially in the central area. This is due to the mixing of the warm air with the cold air that feeds the racks.
CFD study containing hot aisles
We will now proceed to the study with contained hot aisles. For this purpose, a set of vertical panels is introduced into the model so that the hot air is guided upwards through them towards the extraction. The model would look as shown in the following image:
The simulation results show that the system performs reasonably well in terms of air guidance. However, due to the limited height of the screens there is still some recirculation which limits the improvements obtained.
CFD study containing cold aisles
Finally, we will study the effect of containing the cold aisles, so that the air entering the room is directed to the IT equipment without mixing with the rest of the air in the room, which is expected to be at a higher temperature. The following image shows a detail of the model in the confined area:
When the simulation was carried out and the results were extracted, it was found that this alternative provides a much more uniform temperature outside the room. Furthermore, its value is reasonable for the operation of the centre. Containing the cold aisles ensures that the air introduced into the racks comes from the cooling system. This completely avoids recirculations that increase the room temperature.
The final decision on the cooling system will have to consider other aspects, such as the cost of the alternatives or maintainability. However, the use of CFD simulation will provide much more accurate information on the technical quality of the proposed solution. In order to obtain more information on security, the failure cases that should be admissible in the Data Center could be simulated. The time evolution of temperatures in the expected repair or maintenance times could also be considered.
