Study Finds That Humidity Has More Effect On Drive Failures Than Temperature (rackcdn.com)
AmiMoJo writes: A study by Rutgers University and Microsoft has found that hard drives are more prone to failure due to high levels of humidity [PDF] than high temperature. With a view to 'free cooling' data centres (using low external air temperature for cooling to save power), the paper notes that humidity related malfunctions of the driver controller / adapter are the dominant cause of drive failure. The good news is that while the researchers found that high relative humidity was a significant factor in drive failures, "[S]oftware
availability techniques can mask them and enable freecooled operation, resulting in significantly lower infrastructure and energy costs that far outweigh the cost of the extra component failures."
Based on the recent finding, the IEEE has issued new guidelines to augment its previous best practices on disks...some of these are to be expected:
1. the classic practice of gently simmering disks in a succulent gravy is being phased out. gravy increases humidity, and its widely expected a dry rub will not only replace this practice in the future but help flavourful juices reach the surface of the drive
2. RAID arrays should be basted no more than every 25 minutes now, up from 5 minutes. This is partly due to the fact that RAID arrays come infused with a 15% solution of brine. ZFS arrays, as per normal, can be allowed to reduce in their own pools.
3. caching SSD's are unaffected by this finding. A light egg wash or light coat of melted butter toward the end of the partitions is still advised for best performance.
Good people go to bed earlier.
I run a server in my basement and have more trouble with its drives then any other computer I have ever owned. I have left notebooks in hot cars, outside in the sun, exposed them to extreme cold too. But my basement always maintains at least a 50% humidity level even with dehumidifier. Its temp remains between 60F in Winter and 70F in Summer. I noticed one day my router was failing and decided to tear in down, and also realized its soldered board was oxidizing badly which most likely caused the failure. I then wondered if that was the same issues with my server? I do think too little humidity can also cause spurious static charges too so you don't want to not have some. maybe its time to rethink what acceptable is in humidity? I always thought below 50% was OK? But maybe below 40% is the sweet spot?