A liquid cooled server is a server that is cooled by a liquid coolant such as water or mineral oil. Compared to mechanical cooling methods, liquid server cooling is highly energy efficient as the coolants conduct heat much better than air.
What Are the Methods to Liquid Cool a Server?
Liquid cooling is one of the fastest growing technologies in the data center industry, and there are many ways to cool a server with liquid. The method that is best for a facility is based on a variety of factors, including the size of the facility and the density of the racks.
These methods include:
- Liquid immersion cooling. Liquid immersion cooling involves submerging the server in a dielectric coolant. There are two types of liquid immersion cooling: single-phase immersion and two-phase immersion. With single-phase immersion, the coolant never changes its state of matter. In two-phase immersion cooling, the liquid boils, turns into a gas, and eventually becomes a liquid again. Although two-phase immersion offers better energy efficiency, single-phase immersion is more cost-effective and works better for smaller, less dense facilities.
- Microconvective cooling. Microconvective liquid cooling removes heat at the chip level. It involves numerous small fluid jets within compact cooling modules. It is primarily used to improve the performance of applications with the densest compute profile and can cool down even the most powerful processors.
- Direct-to-chip cooling. In direct-to-chip cooling, tubes deliver liquid coolant straight to the chip. The coolant absorbs the heat, and it is then removed. Direct-to-chip cooling is one of the most efficient cooling techniques for reducing data center energy consumption.
- Microchannel cooling. An extension of direct-to-chip cooling, microchannel liquid cooling adds cold plates that directly target CPUs, GPUs, and memory modules. The sealed metal plates spread the heat into small internal fluid channels. This technique is designed to cool a large surface area.
- Oil cooling. Oil cooling involves the submerging of a server rack in mineral oil, which retains heat 1,200 times better than air. The oil is contained in tanks near server racks. The racks are submerged into the tanks where the oil absorbs the heat and is pumped to a passive dry cooling tower.
What Infrastructure is Needed to Liquid Cool a Server?
Although there are various methods to liquid cool a server and each method will require its own set of equipment, there is certain infrastructure required to create a fluid cooling loop. This cooling loop enables the heat transfer between the facility, secondary circuits, and fluid used for cooling. This infrastructure should also be designed to minimize fluid volume to reduce the risks of a leak and to mitigate issues that may arise from facility pressure.
The components needed to liquid cool servers include:
- Coolant Distribution Unit (CDU). The CDU provides controlled coolant for immersion cooling, direct-to-chip cooling, and heat exchangers. It creates a secondary loop separate from the chilled water supply, which enables strict containment and precise control of the liquid cooling system. It normally maintains liquid cooling supply temperature above the data center dew point to prevent condensation and maximize sensible cooling. CDUs also supply the heat exchanger used to remover heat from the rack.
- Indoor chiller. A modular indoor chiller provides support for applications that do not have access to chilled water in the data center. Variable speed pumps allow the flow of refrigerant to adapt to changing loads and water temperature is maintained by internal controls that control the speed of the pump.
- Immersion tank. The immersion tank houses vertically mounted servers in a dielectric bath and removes heat by working with the CDU to circulate cooling fluid.
- Heat rejection. Connecting CDUs with liquid-to-liquid heat exchangers to the building’s chilled water system allows for heat rejection. Heat rejection is the excess heat from a cooling system that begins when the cooling tower heat load is reached. It is the total amount of heat energy transferred from the cool side to the warm side in addition to the work carried out by the compressor.
Monitor and Manage High-Density Racks with DCIM Software
High-density racks are great for fitting more compute capacity in a smaller footprint, but cooling and managing them can be a struggle.
Data Center Infrastructure Management (DCIM) software simplifies high-density rack management and enables you to intelligently increase your density without risking downtime.
With DCIM, you can easily identify and remediate hot spots, set thresholds and alerts on power and environmental conditions to know before you have a serious problem, automate power capacity planning to eliminate stranded capacity, get at-a-glance views of the health and capacity of all your sites, visually document all cabling infrastructure, and visualize the weight capacity of each cabinet in 3D.
Want to see how Sunbird’s second-generation DCIM software makes it easy for you to manage a high-density data center? Get your free test drive now!