Friday March 07, 2008 I was horrified. The room felt damp and warm. My finger slid through a fine coat of dust settled on a flat surface. What's going on here? It is supposed to be meat-locker cold and air-tight. This is a disaster for a high-density data-center, also known as the system lab.
We built this lab only 4 years ago, with extra air-condition units. It is still state-of-the-art: carefully spaced racks, inert-air fire distinguishing, sound-absorbing wall-covering, swipe-card access control, power management, the works. We are Sun. We knew how to do data-centers.
We did not. We under-estimated the growth curve, electricity consumption, and cooling demand. The lab over-heated near the end of last year and we had an "oh sh*t" moment. The options are all ugly: turn off some machines and lose the services they provide, move them to a different lab and suffer disruption, or add more cooling capacity and deal with the hell of funding request. Then, someone came up with a brilliant and stupid idea: open the windows. It is brilliant that outside Beijing air was below freezing; it provided effective and economic cooling. It is so, so stupid that it defeated inert air fire control and shortened the equipment life-span by exposing them with unfiltered air. It also stops working when spring arrives.
Nevertheless, we had time to plan for a move. As this drama unfolds, Sun's eco messages repeated louder than thunders. We experienced the pain of being cooling-capacity constrained. Had we moved to cool-thread technologies earlier, we would have averted the pain.
A testimony for our own gospel? Sigh..