[clug] Re CO2 footprint of Searches: Storm in A Tea Cup or Deep Green Issue?

[Alex Satrapa]

On 09/09/2009, at 10:55 , Andrew Janke wrote:

> Well having once plugged an ammeter in line with a HP2824 gigabit
> switch and a few linksys equivalents (out of morbid curiosity) I can
> say that the power usage when the switch was under "load" vs idle was
> not measurable.

The same goes for my bubblejet printer. In sleep mode it consumes 24  
Watts, when actually printing stuff it consumes 25W, and when scanning  
it jumps up to 30W.

However, that does not mean that it only costs 1W to print, since it  
actually costs 25W to print.

You can't just take the "power consumption when idle" out of the  
equation, otherwise manufacturers could trump each other by sticking  
100W bulbs on the switches which only illuminate when no data is being  
transferred: "Hey guys, check it out! You actually *save* power when  
transmitting packets through *my* switch!"

You have to amortise the "cost-to-exist" over the expected lifetime of  
the product and recover that cost per packet transmitted, in addition  
to the marginal cost-of-transmission.

>  In this case load == lots and lots of NFS traffic
> meaning lots of blinking flashing lights.

... and idle means keeping the machine running so that it's ready to  
transmit packets as soon as they arrive.

> So perhaps the argument here about data transfer from Canada vs from
> local is sort of akin to the public transport bus + train thing. So
> long as the infrastructure is there, there is a negligible extra cost
> for one more passenger on the train.

If there are no passengers on the train, the money it costs to run  
that train is a loss. That loss has to be covered by the profits per  
ticket, or else the whole system has to be run as a community service  
in which case the losses are simply sucking the tax wallet dry.

> ... Meaning I would punt (but
> have no real idea) that the cost of moving data about is predominately
> to cover the initial cost of the equipment, undersea cables, floor
> space, cooling etc.

Exactly. I expect that the embedded energy in a network switch  
(refining and shaping the metal enclosure for starters) is far greater  
than the energy it will consume in operation over its expected  
lifetime. You can't write that energy off before you start counting  
watts-per-bit, you have to spread that cost out over the expected bits  
per lifetime. Thus a very busy network router is "greener" than a  
quiet one that uses half the power, simply because the busy network  
switch is doing more to justify its existence.

Alex