Overheard at the April 2007 Monitoring SIG - I

In some ways, the most interesting aspect of the BayLISA Monitoring SIG that’s hosted here at GroundWork is the post-meeting socializing: while the meetings formally end at around 8:30 PM people typically hang around for a couple of hours after that.

For example: if you think _you_ have a distributed environment consider this: one attendee is an IT manager for a major retailer with almost two thousand stores and five servers per store and, obviously, tens of thousands of networked devices from hand-held inventory counters to POS cash registers networked together. One of his challenges is environmental monitoring in each store’s server room: given that each store is designed differently, and the server room locations are different, and that stores are geographically located everywhere from the cold and wet Pacific Northwest to the arid and baking Nevada desert: how does he know if a given server room’s air conditioning has failed?

The brute force answer is to buy network-able temperature sensors for each server room, but with two thousand stores that runs into some big bucks. Our IT Manager’s McGyver-like solution? Turns out that the off-the-shelf APC uniterruptible power supplies (UPS) he was buying for each server room had temperature logging built in — yeah, it’s intended to report the temperature of the UPS, but he determined that by subtracting twenty degrees from the UPS’s reported temperature he could get the ambient temperature of the server room.

Of course, a thousand questions follow, e.g.: are they all the same model APC UPS? Is the UPS’s location in the server room and in relation to the servers critical? Does he have to do calibration for each UPS? Store-by-store calibration? Seasonal calibration? What happens when someone puts a box of down comforters on top of the UPS by accident? What does he do when APC discontinues those UPSes, or redesigns (or even eliminates) the temperature logging feature without changing the external spec? As he rolls out newer servers with the newest processors from AMD and Intel that include temperature sensors, will he utilize those as replacement or supplemental temperature sensors? Etc. etc.

I only got to ask the first question before I was whisked into another conversation. But I’ve also asked him to present about it at an upcoming SIG meeting: they’re the second Wednesday of every month here at our offices in downtown San Francsicso, a block from AT&T Park.