Exploring the Benefits of Liquid-to-Air Coolant Distribution Units (CDUs)

Liquid-to-air (LTA) cooling solutions have a key role to play in helping the data center industry transition to liquid cooling. Growing demand for AI and high-performance computing requires liquid cooling, but facility retrofits and expansions are time and resource intensive, not to mention expensive. LTA cooling technology brings liquid cooling to the rack level by circulating chilled liquid to servers in a closed loop and rejecting heat into the air where it can be managed by existing air-cooled infrastructure. This helps data center managers bring efficient and high-performing cooling technology to their data centers without the hassle of construction.

How Does LTA Technology Work?

LTA technology goes by many names in the industry: LTA CDUs, LTA heat rejection units and LTA heat dissipation units are all common. In truth, these all refer to the same kind of system—one that takes air from room cooling and pulls it through fans over a coil, which then cools liquid going to cold plates. The heat is then rejected out of the unit into the hot aisle as air; this rejection of heat is where LTA systems get their name.

Deploying liquid cooling without relying on facility water offers significant advantages, such as avoiding the complexities of facility water infrastructure and reducing operational costs. This approach allows data centers to leverage alternative cooling solutions while maintaining high performance and reliability. However, in order for data center managers to realize these benefits, they need to make sure their LTA systems are designed to succeed.

Reducing Operational Costs and Scaling HPC Workloads

An LTA system allows the end user to stack racks with more density compared to a solely air-cooled rack. LTA systems can help reduce costs for data centers by compressing racks together and increasing chip density, as well as helping to bring in additional revenue by supporting more chips. Using the power of liquid cooling, data centers can power and cool more chips while managing their energy use and physical footprint. For colocation (colo) data center in particular, this can be a critical technology, as they are able to establish high-performance computing and AI clusters within their data centers. This allows colo data centers to support AI and high-performance computing for some clients without disrupting existing business.

As long as LTA components have features that are conducive to modular design, LTA systems are very scalable. This enables easy expansion as cooling demands grow since additional LTA units can be added to racks as needed.

Sustainability

Liquid cooled chips reject heat more efficiently than air because of the greater heat transfer capacity of liquid compared to air. Engineers can also design cooling systems with liquid that bring the thermal carrying medium closer to the heat source. While LTA systems may not offer as much overall power efficiency as fully liquid-to-liquid devices, they can still drive sustainability in several areas.

First, the liquid in LTA systems is circulated in a closed loop. This system brings the liquid loop close to IT equipment, requiring less energy to distribute the fluid over long distances. Additionally, while water needs to be filtered to maintain quality and occasionally be replaced, LTA systems do not need a constant supply of fresh water. This can help data centers conserve water in addition to using energy efficiently.

LTA systems also help data centers avoid the environmental impacts of construction. Constructing new data centers or retrofitting existing ones to support liquid lines to every rack can be a very water and carbon intensive process. Using LTA systems can help data centers increase chip density without increasing their physical footprint, minimizing the need to take on these kinds of projects.

The modularity of LTA systems also allows localized deployment of liquid cooled chips, enabling scaling and deployment of additional cooling as data centers increase IT workloads. LTA solutions can cool greater IT loads more efficiently and in a smaller footprint without additional infrastructure.

What to Look For

Outside of basic performance, data center managers should prioritize service and maintenance when designing and selecting components for LTA cooling systems. In particular, designers should look for toolless hot swappable features that provide easy installation and serviceability. As data center managers know, once cooling systems are switched on, they need to run 24/7, so pumps, fans, controllers and temperature and pressure sensors should be serviceable and hot-swappable without turning the unit off. Filters should be cleanable and reusable to keep costs down.

Data center managers should also look for pumps, fans, power supplies and fill-and-drain liquid connections to be located on the front of the unit for ease of service and accessibility. This level of serviceability is absolutely critical to maintain reliable liquid-cooled data centers.

Additionally, data centers that expect to scale their high-performance and AI capabilities need to make sure their LTA systems include features that enable easy reconfiguration like flexible hose connections, and top and bottom feed liquid connection points to IT equipment.