Telecoms Exchange Cooling
CEAD was commissioned in 2012 to resolve a growing thermal problem at a major national telecoms exchange central office.
Servers and telecoms equipment were installed within an old building that was originally used for voice call landline exchange. The proliferation of data and mobile communications had resulted in far more equipment being deployed on site and a much heavier thermal loading than before, the existing cooling and ventilation infrastructure was no longer adequate.
Previous efforts to address the growing thermal issues had focused on the application of off-the-shelf chilling units and air conditioners but in an unstructured way. The result was an inefficient utilisation of compute room air conditioning (CRAC) units, server room real estate and energy while thermal loading on the servers and telecommunication switches continued to rise.
The first step undertaken by CEAD involved gathering information about the current situation. CAD data and drawings of the data centre building layout were obtained along with the installed server and telecommunication equipment types, their locations and thermal performance characteristics. The figure below shows a typical server room construction, similar to that studied and analysed by CEAD.
Figure 1 - Typical telecom exchange server room layout
The acquired information and data was used to construct a computer model of the facility that was suitable for CFD analysis and simulation. The images below show the 3D CFD models constructed and used for the study.
Figure 2 - CAD model of the server room
Baseline Simulation Results
Initial benchmark CFD simulation results were compared with known reported problems as stated by the client. From the results CEAD was able to visualise the airflow patterns and temperature field distribution and thus establish where the main problems were and how improvements could be made across the entire data centre facility. The main problem areas observed in the existing data centre layout and configuration included:
- Over-cooled locations
- Under-cooled locations
- Recirculation zones
Figure 3 - Air flow speed and temperature plots
An extensive site survey was carried out by CEAD over several weeks in order to validate the CFD simulation results. Data logging equipment and sensors were deployed throughout the entire data centre to record temperature, airflow, pressure and humidity during typical day-to-day operations.Critical areas with under or over cooling had been identified from the CFD simulation results, particular attention was paid to ensure sufficient data was collected in those locations.
Comparisons were drawn between the collected data and the CFD simulations, the table below contains typical measured results alongside their respective CFD predictions.
Figure 4 - Numerical comparison betwen measured and predicted temperatures
Good agreement was found between the CFD simulation results and the experimental data, small discrepancies of only a few degrees difference were found and the model was considered accurate and representative of the problem at hand.
Both the CFD simulation results and experimental data were presented to the client before CEAD began work to design and evaluate a solution to the problem.
The main objective was to achieve uniform temperature across the whole server room. This was achieved through the use of the verified CFD model as a basis for evaluation of concept designs. The final solution had to satisfy a number of constraints specified by the client, for example some CRAC units, servers and switching equipment could not be relocated. The solution developed by CEAD involved two main types of changes:
- Some equipment, including CRAC units, were strategically relocated to address identified high temperature zones in order to reduce local temperatures to acceptable levels.
- For immovable equipment, the output air flow rates were adjusted to give a good distribution of chilled air flow whereas previously they were simply set to maximum.
Simulations of the modified CFD model showed significant improvement in comparison to the original setup:
- Lower power consumption
- Reduced dT between inlet and outlet for individual equipment groups
- Elimination of hot and cold spots (under and overcooled regions)
Figure 5 - Temperature plots of the server room before and after the proposed changes
The proposed design was presented to the client and discussion entered into to understand the future expansion plans for the facility. Based on the CFD results and discussion, further analysis was carried out to understand the maximum thermal capacity of the building. From this analysis, a prediction was made for how long the building could continue to support expansion of the housed equipment.
Recommendations were formulated and delivered to the client in a complete report.
Over a period of 18 months the client continued to implement changes based on the recommendations made by CEAD.
The client also entered into discussion with the cooling unit manufacturers with a view to have available products that could be upgraded to have controllable higher cooling capacity, without excessive disruption and downtime to the data centre during thermal upgrades.
The client feedback on the work carried out and the final result achieved was very complimentary to CEAD.In particular they liked the approach taken and the logical process used to solve their problem.
The client purchased similar CFD software used by CEAD and embarked on an internal programme to build a thermal management team. Further consultancy work has been awarded to CEAD to assist their team in strategic design and development of data centres with emphasis on thermal management future proofing.