Testing server power in the data centre
Data centre and telecom
Data centres and server farms consume approximately 1.3% of electricity worldwide, and that share is set to rise with demand for data transmission. Incoming power is AC, the equipment runs on DC, and the voltage has to step down to each server's input — every one of those conversions makes waste heat, and the heat itself then needs cooling.
Where the losses are
Two things generate heat: the conversion stages, and the servers themselves. Cooling the resulting load across an entire hall is its own energy cost, and the whole chain lands in operating expense.
The efficiency response has been centralised AC/DC conversion feeding PDUs that distribute to individual servers. NTT Facilities in Tokyo reduced power consumption by 15% with that configuration.
- Global data-centre share of electricity
- approx. 1.3%
- NTT Facilities reduction
- 15%
Where this figure comes from
Both figures are carried over from Intepro's own page, which footnoted them to a source the archived copy does not preserve. Cite or drop before launch — see docs/QUESTIONS-FOR-THE-CLIENT.md.
The server farm as a laboratory
Some operators use their own estate as a test environment, running prototypes and equipment against a range of voltage and current requirements. That needs programmable DC sources rather than fixed supplies.
An auto-ranging supply covers this with fewer units. Because it delivers lower voltage at higher current, and the reverse, while holding its maximum power rating, one supply spans requirements that would otherwise take several — so the number of sources you have to buy, rack, calibrate and store goes down.
What it shows
The efficiency argument and the test-equipment argument are the same argument. Anything that recovers energy or covers more of the operating envelope per unit shows up twice: once in the electricity bill, once in how much hardware you own.