Compiled by Eamonn Ryan
Data centres are the nerve centres of modern enterprises and societies – and they demand cutting-edge solutions for efficient cooling.
Data centre managers like Africa Data Centre, Vantage and NTT offer floor space or cloud services for businesses to remotely set up their data centres. This shift reflects a move away from traditional building-attached data centres toward more flexible and efficient solutions.
The close of 2023 saw an upswing in tender activity for cooling systems for data centres, signalling a renewed urgency to complete projects before year-end. The market, once in a temporary hiatus, appears to be picking up steam. Data centres are a guarded world, with emphasis on security, energy efficiency and adaptability. This requires HVAC&R companies in this dynamic industry to be committed to innovation and energy efficiency, matching the ongoing evolution of data centre technology itself.
THE ROLE OF CONSULTING MECHANICAL ENGINEERS IN DATA CENTRE DESIGN AND CONSTRUCTION
In the intricate world of data centre design and construction, consulting mechanical engineers play a pivotal role in shaping the future of these technological hubs. Jonker Bester, a director at Spoormaker & Partners, explains that the process initiates with a critical document known as the Owner Project Requirements (OPR) – essentially a design brief by the data-hosting project owner, outlining the project’s fundamental needs.
This can vary quite considerably, depending on how clients approach this stage. “They have expectations at the outset – for instance, a hosting space for 300 or 600 racks at 2kW or 5kW per rack. Some clients adopt a streamlined, blueprint design process with consultants having taken the design up to stage three, which then goes to tender among general local contractors.”
However, in a more traditional process, the path involves a meticulous journey by the consulting engineer through stages one to six, encompassing design, tendering and implementation with the chosen contractors. “The OPR serves as the cornerstone, guiding the entire design process, even if the client is not initially formal in articulating their requirements.”
Sometimes it is necessary to extract essential information from clients, for which purpose consulting engineers utilise design brief sheets. These sheets cover crucial aspects like the number of racks, kilowatts per rack, desired footprint, and the need for ancillary spaces such as staging and operator rooms. Bester notes that sometimes the project owner’s own clients will utilise some space and may have extremely specific requirements. The design brief acts as a guiding document, aligning the client’s vision with the technical aspects of the project.
The collaboration between consulting engineers and architects is crucial from the early conceptualisation stages. “Depending on the project’s scale, the mechanical solution can range from a DX CRAC solution for smaller 100kW to 200kW switching centres which wouldn’t require a chiller plant; to more intricate designs for larger data centres. The emergence of containerised data centres is a contemporary approach that is growing in popularity, allowing for efficient testing and deployment, where a DX (direct expansion) solution also works well.”
CHOOSING THE RIGHT COOLING SYSTEM FOR A DATA CENTRE
In the evolving landscape of data centres, the critical decision of choosing the right cooling system is central to the design phase. Andrew Koeslag, sales manager at HiRef South Africa, explains that suppliers play a crucial role in guiding consulting engineers through the myriad of choices available.
“The diversity of available technologies, coupled with ongoing developments in refrigerants and compressors, presents a formidable challenge for engineers tasked with designing data centres. The dynamic nature of the industry, particularly with the global phase-out of high Global Warming Potential (GWP) refrigerants, adds an additional layer of complexity.” Stefano Migliore, executive director of HiRef South Africa, adds that engineers often find themselves relying on suppliers to stay informed about the latest technologies and refrigerants.
Establishing and maintaining relationships with consulting engineers consequently becomes crucial. During his visits to consulting engineers, Koeslag says he encounters instances where professionals were unaware of the current refrigerants being used in various applications, including computer room air conditioning units (CRAC units) and chillers.
This highlights a potential knowledge gap that suppliers can help fill, offering essential information to consulting engineers grappling with the evolving landscape. Koeslag emphasises the importance of considering new refrigerants. Specifically, for chillers, options include R-454B and R-513A and for CRAC units, R-513A or even R-1234ze. “These synthetic refrigerants, though not natural, are currently more practical alternatives, considering factors like safety and cost and their low GWP. While the industry acknowledges the benefits of natural refrigerants, their adoption poses challenges. Explosive options like butane and propane are A3 classified refrigerants requiring further attention on their applications, while ammonia is not considered viable in the data centre industry. Carbon dioxide (CO2) emerges as a promising natural refrigerant, with advancements in CO2 chillers, albeit at a higher cost compared to synthetic equivalents.”
The industry faces a future where even current hydrofluorocarbons (HFCs) are set to be phased out. In Europe, discussions about their complete phase-down by 2050 are well advanced. Considering many systems are imported from Europe, this necessitates a shift toward natural refrigerants, but their widespread adoption is still in its infancy in South Africa. When queried about HiRef’s involvement in researching natural refrigerants, Migliore affirms ongoing research and development (R&D) efforts. The company is actively engaged in developing equipment that supports these new refrigerants and there’s relentless R&D undertaken at the company’s proprietary laboratories in their Padova-based manufacturing facility.
Koeslag underscores the importance of adaptability in choosing a cooling system: “With a variety of refrigerants, we cater to diverse heat pump requirements, reaching up to 120°C. This is especially relevant in Europe, where a shift from gas boilers to heat pumps is underway, aligning with the broader push towards sustainability.”
Another concept popular in Europe – in some cities even mandatory – but not likely to be seen in South Africa due to the absence of widespread heating infrastructure, is the utilisation of excess energy from data centres for district heating. “Through heat pumps, this surplus energy can be redirected to local industries or district heating systems.” Shifting the focus to the data centre’s heart, the rack units, Koeslag emphasises the prevalent use of R-513A as a refrigerant.
However, a notable trend in South Africa is the increasing adoption of dual-medium units. These units seamlessly integrate chilled water and direct expansion (DX) systems, providing a redundant solution for uninterrupted operation. “Data centres operate on tiered systems, ensuring N+1 redundancy. This means having a base load plus an additional unit for backup, safeguarding against failures in chillers or CRAC units and maintaining uninterrupted operations.”
HiRef develops and tests all its software in-house, adding a layer of proprietary control philosophy. The control system is the brains behind the cooling units, contributing to efficiency, safety and overall system effectiveness. HiRef offers ‘HiNode’, an energy management system designed to offer full visibility, management and optimisation of cooling system operations. It complements building management systems (BMS) by focusing specifically on the cooling system’s efficiency, offering a comprehensive approach to energy management, and acting as a more sophisticated layer that manages and optimises the cooling system’s energy usage.
Migliore notes that there’s a high level of security at data centres, to protect privacy and the information stored therein; strict access control protocols are exercised to restrict access from outsiders, and even technicians require prior accreditation. This is an indicator of the critical nature of the operations within. Even some banks’ data centres maintain a low profile, opting not to brand their buildings. “These measures are not only to safeguard against physical breaches but also to maintain the
confidentiality of the information stored within.”
HiRef’s pursuit of energy efficiency underscores the delicate balance between minimising the cooling system’s footprint and optimising efficiency. “The challenge lies in reconciling the demand for compact systems with the need for efficient cooling, often resulting in a tug of war between conflicting requirements. Data centres tend to be measured in megawatts rather than square metres, there being a correlation between computing power and heat generation. For every kilowatt of computing power, an equivalent kilowatt of heat is produced. This necessitates robust cooling systems, often discussed in the context of larger coils for better efficiency but a larger footprint,” says Koeslag.
IT SYSTEMS NEED CLEAN POWER
HVAC systems play a critical role in ensuring the proper operation of data centres. By maintaining the correct temperature and humidity levels, HVAC systems are necessary to ensure the protection of IT equipment from damage caused by high temperature, dust or corrosion, and to enhance the reliability of data centre operations. For instance, harmonics in electrical systems can have serious consequences, including reducing the life of equipment, causing stress on cables and equipment, and degrading the electrical system over time.
Harmonics are generated by non-linear electronic loads and can cause apparatus overheating, increase power and air conditioning costs, and reduce system efficiency in instances where there are hundreds of EC fans and LED lighting within a data centre. To address this issue, Ziehl-Abegg has developed an active harmonic filter that measures harmonics fluctuation and cleans the power supply to prevent feedback, says Marius du Plessis, MD Ziehl-Abegg South Africa. Typically, an air handling unit (AHU) has multiple fans due to the requirement for built-in redundancy.
“OEMs will provide us with specifications such as a need for air flow at 30m3 a second at 800 pascals. Our system takes into account a worst-case scenario, requiring primary, secondary and sometimes HEPA filters, and maintaining the pressure of the interior a few pascals above the outside – to keep out dust and insulate the building. “The fans are equipped with EC motors and built-in speed control, which allows for precise control of the fan speed and airflow. Ziehl-Abegg’s equipment can measure the flow and control the fans to the desired setpoint, ensuring that the IT equipment is operating at optimal temperature and airflow,” says Du Plessis.
To overcome the issue of dirty filters, Ziehl-Abegg designs the fan wall to accommodate the maximum dirty-filter pressure drop. The fan wall is designed to handle different types of filters, such as normal and HEPA (high efficiency particulate air) filters, which remove the smallest particles. By monitoring the pressure drops of fans, it is possible to determine when to do maintenance on or replace with clean filters. Ziehl-Abegg’s fans are mounted in a fan wall that is part of the AHU manufacturer’s design. The fan wall box is designed to accommodate multiple fans, which are arranged in a fan wall which can have multiple fans in a unit, depending on the required flow rate. Pre-filters are installed before the fan to remove dust and other particles from the air.
The fan sucks the air out and pressurises the system behind it, which usually contains cooling coils supplied by the OEMs. “Another way of doing it is to control the pressure downstream – after the filter, fans and coils – whereby it is possible to monitor and maintain the desired pressure setpoint, say 500 pascals, even if the filters get dirty. In that event, the system will increase the air flow to achieve the 500 pascals set point.” Du Plessis describes the various best practices to consider when designing an AHU for HVAC systems. “The design is critical to ensure that the HVAC system is operating at optimal efficiency. The cooling racks are designed to accommodate different types of fans, such as those that extract heat and vents that bring in cool air.”
Different data centres use different HVAC systems, such as air-cooled systems, liquid-cooled systems, and hybrid systems. The choice of HVAC system depends on the size and type of data centre, the amount of IT equipment, the ambient temperature and humidity, the desired temperature and humidity levels, and the budget. “Our fans are maintenance-free and have built-in diagnostics that measure the air temperature and motor temperature. Vibration sensors are also built into the fans to detect any issues. The EC fans work by receiving an AC supply sine wave, which is then rectified into a DC voltage. The brushless DC motor is highly efficient and quiet.
It operates at the top efficiency level, IE5.” Noise is a significant issue in data centres. “To address this, we use biomimicry to design our fans. We look at nature and design fan blades/impellers that make as little noise as possible while providing the exact airflow required. Our fans are so quiet that you can barely hear them, even when they are running at high speeds. Some blades are serrated (imitating owls’ wings), which helps to break up larger vortices into smaller ones, reducing noise.
This design of centrifugal plug fans on the other hand is inspired by the humpback whale’s tubercles on the leading edge of the flipper, which serves the same purpose of reducing noise and improving efficiency. These fans are used in AHUs for data centres. We apply the same design principles to our repellents, which are also designed to be quiet and efficient.” Energy efficiency is a particular factor in South Africa, especially given power outages are common. Many people are turning to solar power to address this issue. Du Plessis adds: “Our fans are designed to ramp up power slowly, which is important for systems that use solar or wind power. If you put full power into a normal AC motor at startup, it can cause a sudden spike in power, which can damage inverters and other components. With EC motors, the power ramps up slowly, allowing the solar or wind system to generate the extra power required. It takes up to 20 seconds (configurable) for the base motors to get up to speed, which is important to consider when designing a system.”
A BMS controls the temperature and humidity and requests more cooling from the AHUs. “Best practices and infrastructure maintenance are essential to ensure that the AHUs meet our minimum best practices for installation. The plug fans that go into the units need to be sized correctly to ensure optimal performance. We work with consultants to determine the appropriate size and number of fans and build in redundancy as required for each unit. If the consultants don’t provide this information, we recommend the minimum size required to ensure optimal performance. Reducing the size of the fans can reduce efficiency and increase noise levels.”
The BMS system means, via a Modbus or other digital protocols, one can look at all the information on a phone or use a cloud system. From the moment a fan starts one can monitor the hours that it runs as well as the parameters. “You can track the trend as the fan gets older: increasing vibration and temperatures; thereby determining its lifespan and do preventative maintenance before anything breaks down. The bearings’ lifespan is critical – in a worst-case scenario it should be about five years. But we have had fans that have been in use for 15–20 years, depending on their environment,” he says.