catchtwentytwo said:
Some times I wonder if we're making things too complicated.
The most reliable power systems are the simple ones. I think the design of the Piller units is excellent, as they are really simple, and there is less to go wrong than the traditional arrangement.
Heres some things I know:
Any emergency power system that relies on paralleling generators will go wrong one day. Add bonus points if the cooling systems are powered from the output busbars, as then you can be certain that when a genset does hiccup whilst starting and thus the gensets fail to syncronise, the diesels will all shut down from overheat before you crack the syncronisation problem.
So divide and conquer; if the load is two big for one genset (over 2MW) then run multiple emergency power systems, rather than trying to parallel it into one big system.
Electrical engineers like to put breakers and isolators where they aren't needed, because thats the way its always done. Every additional element is a risk of failure. Use the absolute minimum number of things between source and load.
If there is a reasonable cable run between the UPS and the final PDUs, then put static switches right by the PDU, and dual supply the switches. Cyberex make the best (read - most reliable, and longest serving) static switches, and will build a combo static switch with PDU, which is the best approach.
If the data centre is biggish, use dual UPSs and dual distribution, and make sure the IT guys know to plug their stuff into both distributions, so if one does go the IT stuff keeps going. This is a 'Tier 4' technique, and gives the advantage you can work on power kit (and specifically the UPS) whilst keeping the centre running and protected against utility power failure.
Finally, the most important service is cooling. Power to the IT equipment is important, but keeping the cooling running is more important.