Re: Sshh!
I have two anecdotes from the same site.
The first is that one of the supercomputers I looked after was a weight of 3.5 metric tonnes per rack when filled with water, and like Xairan's story, had a steel frame built under the 6 or 7 racks (depending on which of the two clusters you were talking about). Made working under the floor very difficult, although these systems were pretty reliable, and did not require much under-floor work, if you discount the time when cavitation in one of the Water Conditioning Units (think pumps with heat exchangers) caused a serious water leak that had to be mopped up. Interestingly, the way the system was built allowed the faulty WCU to be isolated and replaced without having to turn off the rack it was servicing.
The second story is for a rack of one of the preceding systems, one I did not look after, which was being wheeled along a corridor which had a suspended floor while it was being removed.
They had checked the static weight capacity of the floor, but not the rolling capacity. Consequently, as the rack was being moved, one of the support posts shifted, meaning that all four tiles that had a corner supported by that post dropped, and the rack fell on to it's side, blocking the corridor.
Fortunately, there were only minor injuries, but when the heavier racks that I was talking about earlier were both delivered and removed, the moving company put down 8'x6' heavy duty ply board, and then strong aluminium alloy sheets on top to spread the weight across more tiles and support posts to get them to/from the strengthened area of the floor.
Interestingly, one of the alloy sheets they used was the wrong grade, and when the first rack went over it, it creased the sheet very nicely where the wheels on the rack rolled over it.
Just goes to show that you need to do your homework before deploying these exceptional systems.
Biting the hand that feeds IT
