Auto-sleep modems

[Davey]

The Tim Hortons coffee shop chain in Canada built its reputation on fresh coffee. If it's older than half an hour since it was made, it's thrown away. I've seen it happen.

[Theo Markettos]

Think about it for a minute. Your 'light chattering' is a few packets per second. Each packet takes a few microseconds to process. That means

99.99% of the time your router is idle. That's a useful power saving if acheiveable.

Consumer routers seem to be designed like 1990s PCs - all the packet processing is in software, usually Linux (since it's free). That means we have to have the full CPU and OS awake to process a packet.

OTOH pro routers have a hardware packet path, with software action when necessary. The fast path is power and speed efficient. If there's no packets in the packet path, it's turned off. If the software exception processing isn't necessary, the CPU is stopped. The CPU only needs to wake when an exceptional condition happens, or housekeeping is required. Additionally the network links might be put into a low power state until someone wants to send some traffic.

If this architecture were to trickle down to consumer routers, we would have routers that slept all the time there was no packet to process - not on an idle-for-1-hour level, but on an idle-for-1-millisecond level.

That would be a significant power saving, if the manufacturers could be convinced to do it.

Theo

[spuorgelgoog]

That's handy as people (hope to) move to fibre to the premises and VoIP.

They could save quite a lot if they didn't have so many lights on.

Owain

[Dave Plowman (News)]

Not sure that would make any sense if attempting to save energy overall. A battery isn't an efficient way of storing energy. But it is possible to make a very efficient low voltage supply from the mains - provided you don't spoil things with lots of LED indicators, etc.

[The Natural Philosopher]

used lots of HP printers that go to sleep. But they do wake up when talked to. So fine.

[The Natural Philosopher]

My HP used to wake up every half hour or so and do a warm up and calibration run.

[The Natural Philosopher]

It will come about as the demand for lower power kit becomes more and more voracious.

Its incremental stuff.

Already power on most of my kit is well down from a decade ago.

(loving the SSD)

[Martin Brown]

It doesn't but the crap design of its PSU means that it draws about 10W from the mains to do about 0.3W of useful work running the LED display.

The realtime clock function itself can be done on microwatts (as can a passive LCD display).

[Mike Tomlinson]

En el artículo , The Natural Philosopher escribió:

Funnily enough, I read an allegation today that SSDs use more power when writing data than does spinning rust. Can't say I find it very credible but it would be worth digging for more information.

[Chris Bartram]

Recent office-grade HPs can be set to sleep overnight and auto-wake during office hours to run the calibration etc.

[Brian Gaff]

How does this help anyone? Surely those still using copper wire connected systems have to stay alive to negotiate the fastest speed dynamically over time.

Brian

[Scott M]

Ah, but multiply a few watts by 23,000,000 homes and it's a worthwhile saving in the grand scheme of things.

I'm generally against governmental ninnying, but I have never understood the majority of the population's ability to pee good money up the wall in a completely unnecessary way.

When Windows used to shut down to a "You may now turn off your PC" screen, I worked in an office where someone left it at that screen all night, every night. I didn't know the bloke, but this was a decent electronic engineering firm so he can't have been a fool, but God knows what sort of mental processes go on that they think that's better than leaving it on and letting the monitor go to sleep. Or just turning the bloody thing off.

[Tim Watts]

Few hundred MW ?

As big as the figure seems, it is actually completely irrelevant at a national scale :-o

OTOH 10W waste a day = 2.8p / day and for half the year the heat is not actually wasted as it goes into the house.

[tim.....]

don't McD's have a similar offer with their burgers FWIW

tim

[tim.....]

most places that I work I am in the very small minority in turning my PC off when I go home

most of my colleagues leave theirs on 24-7 (though obviously they all have screens that go into power save mode automatically)

tim

[The Natural Philosopher]

I don't write to it that much...

[T i m]

My Windows Home Server (V1) does that. The last client (Win / Mac) online, going offline allows it to go to sleep and the first client to appear wakes it up again. It won't go to sleep if there is an unfinished torrent, or if the network usage is above a certain threshold (Dropbox could be uploading a big file etc).

According to the stats it's been up 1448 days and in that time cost

1009 kWh and saved (by auto sleeping), 429 kWh. ;-)

Cheers, T i m

[polygonum]

Does that include spinning up or does it assume an already-spinning drive with heads in the right place? Indeed, there are so many ways of questioning the claim. I guess what really matters is whether over a long period of time (hours, days, weeks?) the energy balance is in favour of SSD or disc. (Then start arguing about longevity, energy used in manufacture and disposal, etc.)

[critcher]

surely the main problem with modems is the 10 day renegotiation with the gubbins in the telephone exchange.

[Theo Markettos]

While the radio channel is dynamic and changing, you don't need to analyse it all the time. I don't know what the analysis needs, let's say it's

100ms. If the link is idle, turn off the modulation (which will save a lot of power in lighting the kilometres of wire). When you get a packet, turn on the modulation and transmit using the old link frequency response that you previously measured.

Make sure you renegotiate the response every minute, or whatever is necessary, even if there's no traffic - and if you get packet loss force a renegotiation. That will also detect line faults.

I suspect the challenges are in bring links up and down and negotiating in the shortest time so you can spend longer in idle, but those are things that manufacturers can improve. There might be a slight impact on latency, but it would only affect the first packet that comes in and forces link bringup.

Theo