Why long-term accuracy is important in data center sensor selection 

For a relatively small investment cost, HVAC sensors can make a significant difference to a data center’s energy consumption and power usage effectiveness (PUE). Sensor data sheets may all quote similar accuracy levels, but these figures do not tell you how accurate the sensors will be in five years or even ten. Some sensors drift quickly or significantly, whereas others have excellent stability, maintaining their accuracy year after year. In the following article Anu Kätkä from Vaisala explains why long-term sensor reliability is so important.
 
Accurate measurements are crucial 

Data centers are incredibly energy intensive, so even small reductions in energy consumption translate into big savings. Accurate, reliable sensors provide measurements for the control system, so they’re essential to ensure the building is not over or under cooled or heated. This optimizes energy use while protecting IT equipment from damage and costly downtime. The control of HVAC systems and processes is only as accurate as the measurements.

 

Accurate measurements can also help the achievement of better indoor environment quality, improving the health of employees, as well as increasing the sustainability of operations. Conditions can vary across a facility, so the sensors need to be accurate even in changing or extreme conditions. When buying sensors, data sheets for similar products can be compared, but they do not usually provide any indication of stability over time. It is normal for sensors to drift, but there is a great deal of variation in how fast or far this drift happens. Some products will still be accurate in a decade, whereas others will drift almost immediately. 
 
The risks of inaccurate sensors 

A low-quality sensor might break down faster, requiring maintenance or even replacement. However, a more serious risk is that the sensor might have drifted and be operating inaccurately without the user being aware. ‘Drift’ means that the measurement is actually no longer a valid input for the control system, and this can be difficult to notice unless there is a visible consequence, like equipment being damaged because optimal conditions were not maintained. In the case of data centers, it could also mean that the energy efficiency is not optimal. Energy bills might be too high, which is not easy to spot. Choosing a stable, accurate sensor builds trust in the measurements in the long run; delivering the best possible input to the control system, so that optimal indoor conditions can be maintained in a reliable manner.

 

Data centers consume more than 1% of global energy, and up to 40% of that consumption is needed for cooling and air conditioning. Good HVAC measurements can reduce consumption, while inaccurate measurements can cause it to increase. For example, false high temperature readings can result in excessive cooling. This means more energy is being used, resulting in higher costs and CO2 emissions. False low readings might result in insufficient cooling, which increases the risk of server malfunctions and can reduce server lifespans. Similarly, if air is too dry it can cause short circuits in IT equipment, and too much humidity can cause damage from condensation. Although tolerances have increased, the consequences of inaccurate measurements and unreliable building systems can be dramatic. 
 
Small inaccuracies lead to large costs 

Vaisala carried out a case study to simulate how much more energy would be used if there was just one-degree measurement error, causing too much cooling. The results showed that a one-degree error increased energy consumption by more than 8.5%. With the size of data centers and assuming the same deviation is happening over a decade, this seemingly tiny error translates into many millions of euros of extra costs. Knowing that all sensors do drift over time, some quickly or significantly, whereas others have excellent stability, this deviation should be considered a real risk. 

 

Aside from financial savings and sustainability considerations, uptime is also worth considering. In this line of business the servers cannot be unnecessarily down. Uptime of 99.999%, as many high-level data centers want to promise, translates into just five minutes of downtime across an entire year. The higher the tier of the data center, the more catastrophic an uncontrolled downtime will be.
 
Invest in long-term thinking 

The more critical the building, the more important the sensors. Minimizing the total cost of ownership is different from minimizing the cost of investment. High-quality sensors will last 15–20 years so it is worth purchasing those that maintain their accuracy over time, and are designed for effortless calibration and maintenance. This will help to keep the business running optimally, saving energy, and helping to achieve better PUE.  

 

Cheaper sensors will require more maintenance and need replacing more often, but the invisible threat is that they may drift without users even realizing, leading to sub-optimal indoor conditions and poor energy efficiency. It is wiser to think long-term over the sensor lifecycle and to understand and compare the total cost of ownership – when it comes to stability of measurement, the comparative unit cost is only a small part of the story. Real lifetime value is obtained when a decision is made to invest in reliable, high-quality sensors that maintain their accuracy long-term. 

 

For more information on the importance of long-term accuracy in sensor selection, contact Vaisala's team of experts.