Real Time Temperture Monitoring, Logging.

The nifty temperature sensor (mentioned in a previous post) showed up on Thursday, and was installed at our data center today. This is when the geek in me really shows. Here's a device that combines an embedded server (probably Linux-based, although I haven't checked for certain) and a data acquisition component (in this case, one or more temperature sensors). The result is a compact, rack mountable device that will display real time temperatures for multiple locations. It will also log the temperatures at a variable time interval, which defaults to every 5 minutes.

I've rack mounted the device in the top of the cabinet, and run the second sensor to the bottom (a distance of approximately 72 inches). The important thing to know is that the cooled air flows through the raised floor; therefore, the differential between the two temperature sensors will represent how much heat is added by the equipment in the cabinet (i.e., servers, disk arrays, switches and firewalls). So what are my expecations?

First, the cool air supplied through the raised floor should not fluctuate much. Fluctuation in the temperature of the supplied cool air would indicate a failed cooling unit or thermostat.

Second, there's a possibility that a measurable differenital in temperture could coincide with the daily network traffic patterns. Now that would be interesting. Also, there could be a spike in temperature during the scheduled backup procedures when the multi-terrabyte nearline array is running at full clip.

The small sample taken following installation today indicates an average differential of 22^o F. That puts the temperature in the top of the cabinet at 82^o F. (Does anyone other than me think 82^o F is a bit too high?) I realize that the upper tolerance for the equipment is 104^o F, and that I'm measuring the exhaust temperature (not the temperature of the air flowing through the equipment), but additional equipment planned for the cabinet will push the temperature even higher.

As with most new information, there are new questions to consider. Additional temperature sensors might need to placed in the cabinet to test for air circulation patterns. The cool air should be flowing from the front to the back, in addition to flowing from the bottom to the top. If this isn't happening, the equipment could be starved for cooled air.

I'll post graphs once I have a large enough sample size to work with.