3-4MW is about right for each of the datacentres we have servers in, what fraction of those MW are sipped during the peaks of the sine wave? How many "big" UK datacentres? 40-50 maybe?
- Vote on answer
- posted
10 years ago
3-4MW is about right for each of the datacentres we have servers in, what fraction of those MW are sipped during the peaks of the sine wave? How many "big" UK datacentres? 40-50 maybe?
Just about ANY electronics tends to have an SMPS strapped on it. And they are ALL peak clippers.
It's a huge change from the days when the load on the grid was largely resistive (lighting) and slightly inductive (electric motors).
Peak currents are much higher than the average..the current waveform is far from sinusoidal. This leads to far higher losses in the resistive parts of the grid.
Each.....
In 2011 the total UK datacentre power was estimated[1] as 6.4GW with
6.7% growth P/A. Assume half that's cooling, so relatively smooth inductive load, and the other 3.2GW is consumed during 4ms out of every 20ms cycle, which makes it equivalent to 16GW peak, or 2/3 of the UK's average base load ... still insignificant Harry? [1]
Yes, but you're a tosser and I actually know what our data centre draws.
your data centre draws over ten megawatts?
Trivial. And most of it is in ventiilation anyway, nothing to do with electronics.
Twaddle.
Most of it goes on cooling/ventilation not DC driven cevices. I was reading only lately that data centres are being moved to colder climates than ours to get round the energy consumption /cost problem. Finland & Canada being favourite.
strap a current probe on the mains and plug the probe into an oscilloscope and the mains into any item of electronics.
just DO it harry.
Or just THINK for a moment how a sine wave and a rectifier can charge up a capacitor ins a standard single phase pridge rectifier and capacitor arrangement, such as every power supply that produces DC off the mains uses, when for the greater part of the cycle the diodes are reverse biased and cannot conduct any (significant) current at all.
" Power consumption in UK data centres was estimated to be 6.4GW annually."
Interesting. What is anuual power consumption? Energy? GW isnt energy...
twaddle.
even basic schoolboy theory will tell you that a heat pump can remove
3-4 times the power it consumes in terms of heat, so the natural ratio would be 75% -80% devices producing (unwanted) heat and 20-25% running the aircon.
The far more relevant transition going on is the massive move to virtual servers where one server can replace up to 50 (or more) servers running different applications on different operating systems with a single piece of hardware, and the recognition that 'megaflops per watt' is a figure CPU designers should be optimising.
It is possible to remove all the flux, even from a transformer
In article , harryagain scribeth thus
Then why do they cause waveform distortion and why does the mains get distorted?..
Even with very light transformer loading's?..
In days of yore I had to manage a TV repair shope and in those days we used to have a lot of the Philips G8 chassis sets which used a single thyristor in the power unit.
However from time to time we used to have a rather odd complaint that when the telly was switched on the lights would flicker and when TV was switched off all was well.
Invariably it was caused buy a duff high resistance joint in the fuse box or incoming mains supply.
That HR joint could cope with resistive loads OK, but the odd half wave demand of the TV caused it to play up and cause the effects alluded to above. We had known duff meter connections that would supply the cooker etc but when the TV was switched on then the flickering started!....
I have known that up to the 440 MHz band!...
There is a move towards putting data centres where its cold so you can avoid the cost of cooling. Some are being sited in Iceland where they can be hit by volcanoes!
I wonder if it will be like Japan, have standby generators 25 miles away from a nuke melting down and deciding not to use them?
That does not really stand up to even a cursory analysis though does it?
Think about how efficient a modern air conditioning system is... (consider the meaning of COP in these circumstance)
Now think about the energy that you are needing to pump away, to maintain an equilibrium temperature in your data centre. Where does it come from?
Hopefully, it should now be obvious that a cooling power load can't represent "most" of the energy consumption.
the speed of striking of a fluorescent is..pretty high!
Yes, but only if by "twaddle" you mean, "yes that is a pretty good description", then you are right.
Here I drew it for you:
The waveform shows voltage at R2 and current through it.
Actually I think taking a figure of half for cooling is quite fair, one of the datacentres we use has active and passive cooling, and they only use the active cooling at times of year that demand it, that one in probably 25-30% cooling.
There certainly is an element of that going on.
HomeOwnersHub website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.