Taming the Coil-Icing Beast
By David Inniss, SynapSense Director of Product Marketing
As manufacturers of servers and IT equipment continue to improve computing power and reducing equipment size, an increasing amount of unwanted heat must be rejected from the data center. There are several approaches for achieving this heat removal—which approach to use depends on the mechanisms used to transport heat energy beyond the walls of the data center. To that end, this blog focuses on Direct eXpansion (DX) systems. Inherently, DX is thought to be synonymous with “air-cooled system”, but in reality, any system with an evaporator and a refrigerant flow can be termed “DX”.
Recent trends suggest the increasing popularity of DX cooling systems as the best approach for heat removal from data centers and computer rooms. The ease and lower costs of installation and maintenance have contributed to this popularity as part of original building design or as a retrofit initiative to add load to an existing space. This is especially true for small to medium sized facilities trying to leverage the advantages typically associated with DX heat rejection.
When combined with the general lack of visibility into thermal conditions in the data center, perceptions about the heat sensitivity of IT equipment and a strong desire for 100% uptime are two of many factors contributing to widespread overcooling. In fact, excessive overcooling is the norm and is often viewed as an “insurance policy” against temperature-based server failures. However, increasing energy costs, decreasing operating budgets, and the prioritization of “green” initiatives, have placed focus squarely on methods to eliminate energy waste associated with overcooling while retaining resiliency and the “warm and fuzzy” assurance for data center operators.
An alternative to DX Cooling is the traditional Chilled Water system. Reducing energy waste in chilled-water cooling systems relies primarily on the ability to lower fan speeds. The Fan Cube Law mandates that lowering fan speeds by 20% can potentially reduce fan energy used (up to 50%). In fact, lower fan speeds have secondary effects. Slower moving air can lead to better heat transfer and less bypass air and turbulence. In turn, less bypass air and turbulence lead to better stratification and higher return temperatures, further improving efficiency.
The story with DX cooling systems is quite different. Simply reducing the fan speeds without clear understanding as to the impact can lead to undesirable outcomes and risks of damaging or destroying a unit. Changing the airflow across the evaporator coil in a DX unit without understanding and carefully evaluating such changes introduces the risk of ice forming on the coil (coil icing), ultimately causing system shutdown. Certainly, this blind pursuit of energy savings by attempting to address overcooling introduces the same threats to resiliency that overcooling sought to mitigate in the first place. This is not good!
Coil icing happens quickly and conditions can degrade from acceptable limits to critical in less than 10 minutes. Furthermore, once ice begins to form, the evaporator goes into a tailspin that is difficult to recover from without shutting the unit off. How does this happen? Because of differences in thermal conductivity of aluminum and ice, the formation of a small amount of ice prevents the coil from absorbing heat, which causes the coil to become even colder as the ice spreads within the unit. Supply temperatures will continue to increase as the ice formation blocks all airflow.
These resiliency risks can preclude aggressive pursuit of energy savings from DX units unless (a) more visibility to the internal conditions in the unit is employed and (b) there is intelligence that identifies the conditions that may lead to the icing effect and avoids them. While newer units may provide such visibility, retrofit of older units with VFDs should include a mandatory step of applying the appropriate instrumentation and “smarts” to identify icing conditions. The old adage still applies, “You cannot control anything (safely) without visibility into its performance.”
One approach to addressing the identification of these conditions is the SynapSense ThermaNode™ DX. This device offers insight into internal operational status and thus provides the ability to predict the waking of the “coil-icing beast”. When combined with the Active Control™ module of the SynapSense Data Center Optimization Platform, the pursuit of energy savings no longer needs to happen at the expense of resiliency.