Monitoring temperatures in data centers: Planning the sensor system correctly

Temperatures in data centers can be reliably assessed if measurements are taken at the relevant locations and the data is incorporated into a defined operational process.

A single room-wide measurement may overlook local hotspots. Therefore, temperature distribution, humidity, and airflow must all be included in the monitoring concept.

For specialist planners, building automation systems, and technical operators, the solution begins with three questions: Where do critical conditions occur? Which sensors make them visible? And who responds to a deviation? This article presents a comprehensive approach, from the measurement point plan to the building management system.

The focus is on air-cooled racks and their room environment. For liquid-cooled systems, additional measurement and monitoring tasks are required within the cooling circuit.

Back to blog

What is a reasonable temperature for data centers?

As a guideline, ASHRAE (American Society of Heating, Refrigerating and Air-Conditioning Engineers) specifies a recommended range of 18 to 27 °C at the air inlet for air-cooled IT equipment in classes A1 through A4. This reference point is important: A value measured at the room wall or in a hot aisle reflects different conditions. The recommendation does not apply universally to every equipment class and every cooling technology. The requirements of the actual hardware in use remain the determining factor. Source: ASHRAE Handbook, Chapter: Data Centers and Telecommunications Facilities.

The recommended operating range, the permissible device range, and the project-specific alarm value serve different purposes. Therefore, a permissible limit value is not automatically a setpoint for continuous operation. Define operating targets and warning thresholds in conjunction with IT and cooling planning.

Furthermore, an unnecessarily low cooling temperature does not cause poor air distribution. Before changing setpoints, temperature distribution, load conditions, and cooling performance should be systematically monitored and documented.

Four Challenges in the Data Center

  • Local Overheating

    A room-wide reading may appear normal, while a single intake area becomes too warm. Distribute measurement points according to the actual airflow paths and critical rack areas.

  • Humidity Without Context

    A percentage value alone does not describe the risk of condensation. Only the interaction of temperature, humidity, and cold surfaces allows for a reliable assessment.

  • Changes to the airflow

    Modifications, changes in load, or a change in the system's operating condition can affect the supply to individual areas. Compare the temperature map with additional diagnostic parameters.

  • Gaps in the alarm route

    Measurement data only protects the system when it is properly analyzed and acted upon. Also check for communication errors, clarify responsibilities, and ensure that alarm recipients can be reached.

A measurement concept begins with operational questions

First, create a measurement point plan showing the rack and room layout, airflow paths, cooling surfaces, and potential leak locations. Assign a specific question to each point: Should a local overtemperature be detected, an air condition be evaluated, or a water leak be reported?

This results in a list of data points that includes the measured variable, installation location, required accuracy, interface, and evaluation. Add the responsible party, warning threshold, and the expected response. Also consider how sensors can be checked, replaced, and added in the event of an expansion.

When selecting a sensor, the key consideration is not the number of functions it offers, but rather the specific measurement task. A room sensor, a duct sensor, and a surface sensor are therefore not interchangeable. You should also determine what information must remain available in the event of a device or communication failure.

Contents

Temperature Measurement Points and Hotspots Humidity and Dew Point Differential Pressure and Airflow GLT and Alerts Common Planning Questions

Identifying Hotspots: Measure Temperature Where IT Components Draw in Air

A room sensor measures the temperature at its installation location. However, it does not automatically indicate the actual air temperature around individual servers. Variations in server occupancy, changes in IT workloads, or an unfavorable mix of supply and exhaust air can lead to local temperature differences.

Therefore, plan measurement points along the relevant airflow paths. For racks with front-to-back airflow, the intake areas are particularly indicative. Measurement points at different heights can reveal whether certain areas of the rack are receiving insufficient airflow. The diagram illustrates this principle; it is not a blanket guideline for a fixed number of sensors.

For each measurement point, document the rack, side, height, and the corresponding measurement task. Additional readings for supply and return air help assess the system's condition. Compare the readings under known operating conditions and recheck the distribution after changes in occupancy or airflow patterns.

To select a appropriate temperature sensor Other factors to consider include the measurement range, accuracy, response time, and a design suitable for the measurement location. A wall location that is easily accessible is not necessarily the right place for every measurement task.

Humidity and Dew Point: Properly Assessing Condensation Risks

Relative humidity and temperature must be considered together. If the temperature changes, the relative humidity may change even though no corresponding amount of water vapor has been added or removed. A single percentage value therefore only partially explains the state of the air.

The dew point describes the temperature at which the air in question becomes saturated as it cools. If a surface reaches or falls below this value, condensation can occur. For cold pipes or cooling surfaces, the surface conditions are therefore also relevant. A seemingly normal air temperature does not rule out the risk of local condensation.

Combined Humidity and Temperature Sensors Depending on the model, S+S sensors are available with analog outputs such as 0–10 V or 4–20 mA, as well as with Modbus. The building management system can calculate the dew point based on temperature and humidity; compatible Modbus models already provide this value as a data point. Please refer to the respective product documentation.

For planning purposes, the measurement location, expected air conditions, potential cold spots, and the intended response are determined jointly. The sensor provides the basis for the measurements. The control or monitoring system designed for this purpose determines whether cooling should be adjusted or a malfunction reported.

Differential pressure reveals changes in airflow

Temperature readings indicate the result of heat dissipation at a specific measurement point. An additional differential pressure reading can provide insight into changes in the air supply. Depending on the system design, this may involve, for example, comparing two defined areas of an enclosure or an air distribution system.

Clearly documented reference points are essential. Note the locations of the two pressure connections, how the measurement lines are routed, and which operating condition is used for comparison. Without this information, the same change in numbers can be interpreted in different ways.

When selecting Differential Pressure Transducers The measurement range, accuracy, and system interface must be compatible. Depending on the model, analog outputs such as 0–10 V or 4–20 mA, as well as Modbus, are available. Also, pay attention to the permitted media: An air differential pressure sensor is not automatically suitable for liquids in the cooling circuit.

Evaluate the pressure trend in conjunction with temperature measurements and the status of the cooling system. A pressure differential alone does not confirm that each rack is receiving an adequate supply. Similarly, a specific Pa value cannot be generally specified as a target value without a system and measurement point design.

From the Sensor to the Building Management System: Measurement Values Must Enable a Response

A comprehensive concept covers the entire chain: sensors, controllers, or building management systems; a higher-level monitoring system, if necessary; and a defined alarm path. It must be determined during the planning phase which system will collect, store, and evaluate data and forward it to the appropriate personnel.

Analog sensors with 0–10 V or 4–20 mA can be integrated into the control system via compatible inputs. If these inputs are already available, they can be a cost-effective choice. Modbus Depending on the device, this allows for the transmission of multiple measured variables and additional data points via a common bus. Check the signal type, as well as the registers, scaling, and units. When comparing costs, consider not only the device price but also the number of inputs required, wiring, and integration effort.

A Modbus sensor does not automatically send emails or text messages. This requires a system component that is appropriately equipped and configured. It is equally important to distinguish between a genuine threshold violation and a missing or outdated measurement. Communication failures must not be mistaken for a normal operating condition.

Check the entire chain of events leading up to the operational response: Is the correct value being read? Is a disruption being detected? Does the message reach the intended recipient? Are the responsible party and the next step known? Test these points before handing over responsibility and after any significant changes.

In the case of a Retrofitting Existing Buildings The availability of inputs, wiring routes, interfaces, and the expandability of the existing building management system should be clarified early on. The solution must be compatible with the existing system and the operational organization.

Monitor condensation and leaks separately

  • Monitor Condensation

    Condensation occurs when conditions on a surface allow dew to form. Depending on the model, condensation or dew point monitors detect such conditions or when conditions approach them. The installation location and operating principle must be tailored to the specific surface being monitored. An alert is integrated into the planned control or fault logic.

  • Detecting Leaks

    A leak sensor detects escaping liquid at the designated monitoring point. This is a different task from determining the dew point. When selecting a sensor, consider the type of liquid, the location, and the detection range. Condensation Monitors and Leak Sensors should therefore be selected based on the specific cause and the desired reporting behavior.

Conclusion: Plan measurement tasks, sensor systems, and operations together

Reliable temperature monitoring in a data center combines meaningful measurement points with appropriate sensors and a proven alerting process. Temperature, humidity, dew point, and differential pressure complement one another when it is clear which question each value answers. Leak detection is added where liquid leaks need to be detected early.

S+S offers a variety of sensor and monitor families for this purpose. The specific model is selected based on the measured variable, mounting, accuracy, and interface. Project-specific availability, operating limits, and system integration should be confirmed using the device documentation before placing an order.

Coordinate technical selection with S+S

Sources and Further Reading: ASHRAE on the Operating Conditions of IT Equipment; Raritan to Exemplary Measurement Points on the Rack; Vaisala on Sensor Requirements in Data Centers.

Back to the Knowledge Blog

Frequently Asked Questions About Temperature Monitoring in Data Centers

What should the temperature be in a data center?

For air-cooled IT equipment in classes A1 through A4, ASHRAE recommends a range of 18 to 27 °C at the air inlet as a guideline. This range is not a one-size-fits-all target value for every system. Check the equipment class, hardware specifications, and cooling design. Room temperature, hot aisle temperature, and the air drawn into the equipment are different parameters and measurement locations.

How many temperature sensors are needed for a server room?

The number is determined by the room layout, rack occupancy, airflow, and monitoring objectives. A single room sensor only monitors its installation location. Additional measurement points are planned to account for local temperature variations. Start by creating a measurement point plan and verify coverage under various load conditions. There is no universal number of sensors for every server room.

What role does humidity play in a data center?

Humidity, together with temperature, describes the state of the air. Another factor relevant to potential condensation is whether cold surfaces reach or fall below the dew point. Therefore, humidity readings should be evaluated in conjunction with temperature and surface conditions. Depending on the model, suitable transmitters can also transmit the calculated dew point to the building management system.

Is Modbus absolutely necessary for monitoring?

No. Sensors with analog outputs, such as 0–10 V or 4–20 mA, can be integrated via the appropriate inputs on a controller or building management system (BMS). Modbus is a digital alternative, for example, when multiple data points are required per device. The key factors are the available interfaces and the measurement task. For alarm signaling, both options require a system component that is appropriately equipped and configured.

What is the difference between condensation monitoring and leak detection?

Condensation monitoring refers to condensation—or the conditions that lead to it—on a surface. Leak detection detects escaping liquid at a monitoring point. These two tasks require appropriate sensor principles and installation locations. In a system, both functions can be useful if cold surfaces and potential liquid leaks are to be monitored separately.