Humidity in the Data Center: The Real Hard Drive Killer
Researchers at Rutgers University found that humidity, not heat, is the single greatest detriment to a hard disk drive’s lifespan in a recently published HDD reliability study. This revised understanding of the ideal conditions for prolonging HDD longevity has the potential to change how data centers manage environmental conditions. The study examined one million HDDs across nine data centers over a period of 18 months to four years for each center. HDD failure is by far the most common hardware failure type at data centers, being responsible for 76 to 95 percent of all component failures. Not only can HDD failures mean lost data, but they also can add up in costs to replace.
The study found that HDDs operating in higher-humidity conditions are substantially more likely to experience disk controller and connectivity errors. The environment exposes the controller board and disk adapter to condensation, which leads to corrosion and eventual device failure. The study reached this conclusion by comparing failure rates to outside temperatures between high-humidity and dry conditions. When the temperature spiked, the humid environment HDDs experienced a substantial increase in failures whereas the dry environment devices did not.
When Enemy Becomes Friend
The Rutgers study doesn’t give a green flag to expose your data center’s storage components to extreme temperatures. The findings imply that there is a warmer temperature range where the extra heat causes less damage to the hardware than it prevents by eliminating humidity in the device proximity.
This study actually built on the findings of Google’s 2007 study “Failure Trends in a Large Disk Drive Population,” which identified a “sweet spot” where HDD operating temperatures minimized the odds the device would break. However, the Google study did not examine humidity as a factor in why HDDs fail. According to the Google study, HDDs have the lowest failure rates when operating between 38 and 45 degrees Celsius, which supports the idea that the higher temperatures combat humidity-related damage. Google’s study also found failure rates are the worst when the devices operate under 25 degrees Celsius and above 48 degrees Celsius. Drives that get too hot will eventually cause more damage from heat then they prevent from humidity at that temperature.
Better understanding about why HDDs fail provides insight into building servers and infrastructure that works with lifespan extending practices instead of against them. Data centers that are located in higher-humidity environments can improve HDD operating conditions by using servers that place the HDDs near the airflow exhaust instead of the intake. Bringing in the cooling air over the CPU and RAM increases the temperature and reduces the humidity, so the air that cools the HDDs is warmer and less humid. Competitively, bringing in the cooler air over the HDDs first and exhausting over the other server components does not have the same benefits. The data also argues that data centers in hotter, drier environments instead of cooler, more humid ones will have less-frequent hardware failures.
Comments are closed