Slashdot Mirror
Internet's Energy Needs Growing Faster Than Efficiency Gains
Electrons may not weigh anything, but it takes some heavy lifting, both literal and figurative, to point them in the right direction. Reader terrancem writes with this excerpt: "Energy efficiency gains are failing to keep pace with the Internet's rapid rate of expansion, says a new paper published in the journal Science. Noting that the world's data centers already consume 270 terawatt hours and Internet traffic volume is doubling every three years, Diego Reforgiato Recupero of the University of Catania argues for prioritizing energy efficiency in the design of devices, networks, data centers, and software development. Recupero highlights two approaches for improving efficiency: smart standby and dynamic frequency scaling or CPU throttling."
You lost me there.
Maybe we should be using this new fangled internet thing to figure out how to come up with more efficient power sources. Then we could expand the internet. Wash, rinse, repeat.
270 terawatt hours! Is that per hour I wonder ?
Actually, the mass of an electron is: 9.10938291(40)×1031 kg :-)
What about the energy offset?
How much energy is consumed by driving to blockbuster, picking up a physical tape that had to be produced and shipped to the store Vs. streaming from Netflix?
How about paying bills online vs mailing an envelope.
I'm not sure what the number is but it may be possible that for every increase in energy 'x' by computers there was '5x' amount of energy saved in other areas???
I'm not goin back to layin in the back yard and waiting for a boob shaped cloud to float by....
Solution to reduce internet traffic by 30%. Make massive spamming an offense punishable by death or life in prison. Set up an international enforcement body. I am 100% serious.
In how long?
Could be 30 gigawatts for a year, 300 megawatts for a decade, 370 gigawatts for a month or even 16.2 petawatts in a minute.
Units matter!
We need a binary protocol and tools to handle that binary protocol.
Yeah, I know, text is ubiquitous, but so this new protocol will be if it is open source.
But a binary protocol will reduce consumption by a large amount.
Recupero highlights two approaches for improving efficiency: smart standby and dynamic frequency scaling or CPU throttling.
Turning a 1 hour task at 100 watts in to a 2 hour task at 75 watts isn't efficient.
If we colonize Mars, it won't be the World Wide Web anymore. UWW?
and plant some trees? then it evens out
Usually when people use stuff, that involves resources.
On the other hand, more and more content is being consumed by mobile devices, which are vastly more energy efficient than desktop computers. Even desktop computers are more efficient than 10 years ago, primarily due to the complete abandonment of power hungry CRT monitors. So the good news is the part that's hard to control, which is the diverse and eclectic individuals who consume the content, are already many times more energy efficient than they were 10-15 years ago.
Better known as 318230.
hard drives suck up the most power
i'm sure the government can make up a tax credit to get people to buy up SSD's
> Internet traffic volume is doubling every three years
Compute power doubles every 18 months. And storage doubles faster than that. Performance per watt has been growing exponentially. Where's the problem?
I looked at the figures, and they are laughable. Right. Good luck using "smart" power management to reduce clock frequencies *between packets*. If you've got that much overcapacity, you can just put an entire node to sleep and redistribute the load.
Granted, I have not read the article in science, but it seems like this got the green-light from editors who are not in the field.
This is like saying gas usage increased faster than engine efficiency when cars became widely adopted. Well duh, more people than ever are doing more things then ever online, and that is going to consume more resources. I would like to see how this curve correlates to user counts and usage rates.
We'll just put all the people in energy pods, and the problem will be solved.
Did anyone notice the meaning in Italian of the paper author's last name is "re-forged recycling"?
Computing is orders of magnitude more efficient efficient than the traditional ways it replaces.
Actually, the mass of an electron is: 9.10938291(40)×10^31 kg :-)
I think you might have missed a minus sign there. Unless the Sun is an electron.
Damn. I want to know what universe you live, in, those electrons are HEAVY.
The sun only weighs ~1.9891x10^30kg, an electron is almost 5.0 x 10^1000 times heavier!
( http://en.wikipedia.org/wiki/The_Sun )
Self proclaimed typo king, and inventor of the bear destroying coffee table (patent not pending).
>>> Electrons may not weigh anything, but it takes some heavy lifting...
Year-on-year the editing gets worse and worse.
What are we, 5-th graders?
They're not photons or neutrinos. So yeah they do weigh something. If I was writing a submission summary I'd check that kind of statement in case I end up looking like (more of a) gimp. ;)
I do not want your cheap brainburning drugs. They are useless for work. And I am a working man today.
.. makes sense, after all, there was no consideration about the actual power requirements, the Internet has been work in progress since its inception. It has grown, but it's never been looked at as a whole in terms of hardware. But to make it energy efficient in a purist sense, there would have to be enforced standards on hardware requirements which would entail ISPs to reevaluate and tweak their setups. This won't be cheap, but I'm sure it will still be cost-effective compared to business as usual.
Internet pron is causing global warming, m'ok? Save the environment, buy print editions of Playboy, Penthouse and Hustler.. or whatever else gets you off.
hours in 3 years: 3*365.25*24 = 26298 h
average power over that period: 270 TWh / 26298 h = 10267 MW
So... like 8 simultaneous lightning bolts to run the internet for a fraction of a second?
Getting tired of Slashdot... moving to Usenet comp.misc for a while.
Government monitoring and storage of all communications of its citizenry has got to have a tremendous carbon footprint. As does all the extra electricity used by Facebook, Google, Double Click et all to track my every move on the internet. How much energy could be saved by simply serving web requests, and not data mining it for government and corporate interests?
Ferchrissake -- get your units right.
I wonder how much does impact does Java and enterprisy XML-based web services have in all that. XML is a cache hog and memory bandwidth hog, never mind a network bandwidth hog. Java has huge runtime costs of abstractions needed for good software design. I think it's time to come up with something where the abstractions' cost is pushed to compile time. You know, something that has been solved long ago in the form of LISP code-generating macros. Sigh.
A successful API design takes a mixture of software design and pedagogy.
But I prefer robust, something that can tolerate the occasional anchor drop, route around censorship, and protect anonymity.
“He’s not deformed, he’s just drunk!”
Killing us softly, entertaining us 'til death pries our illusions from our cold, dead hands.
Wouldn't it be ironic if the world's greatest communication leap accelerates global climate change and throws us all right over the tipping point?!
Let me know when you 2 scholars have finished computing the angular momentum of the internet.
If governments and corporations would just stop collecting all that data on everyone...
I find it odd that the smart standby graphic displays devices spooling up in anticipation of packets. prehaps they are using quantum entanglement devices to instiontannously signal an incomming packet and by the time the packet arrives, the server is spooled up.
Is that what the sun weighs on earth or the moon?
By not having to mail letters, package software in retail packaging and ship to a store or end user, travel to other office locations for meetings, and other spendy endeavors?
Overall the Internet is a huge energy saver.
Or, instead, you could just switch to DC power (15 pct drop) and Solid State Drives.
But that would be ... prudent.
Try not hosting them in hot climates, like the ones you're creating by using oil and coal to power them.
... ensures we're all doomed anyway.
Yeah!
On the sun, obviously ;)
Which of course, raises the question, why couldn't you just bloody SAY "31 Gigawatts" instead of tangling yourself in this foofaral of extraneous time units that you didn't even get right?
The watt is a measurement of power. The kilowatt-hour is a measurement of energy. 31 Gigawatts on its own is meaningless. That incandescent light bulb in your closet is rated for 100 watts, does that tell you how much it cost you to operate last month? The power consumption is useless without knowing how long the device was turned on, and it's easier to say "That light bulb consumed 15kWh last month" than to say "That 100 watt light bulb was turned on for 150 hours last month."
I want peace on earth and goodwill toward man.
We are the United States Government! We don't do that sort of thing.
I wonder how much of the energy being used in data centres is being wasted. And not just as in heat. I worked in one centre and we had one rack of blade servers that wasn't being used for anything but was constantly being left on. I would turn them off and next time I looked they would be turned on again. And they put out a lot of heat into the room. We had another two full racks of blades from another provider that were in constant use and they put out less heat than the one rack sitting idle.
Hopefully that has been changed there but I'm sure that it isn't the only place in which there are unused servers sitting turned on and left to idle 24/7.
I just finished writing an article on 16 ways to save the planet. Number 8 was to institute a efficiency standards. Design a moving goal post to keep pressing the issue in a sort of energy efficiency Mores Law. Currently our brains are a million times more efficient than the computers we run and at the same time are a million times more powerful! If we press the issue and put money into it we can build the technology to get our computers to match the efficiency of the human brain. There already has been several designs proposed to make this happen including using old analog types of computer designs instead of digital ones which are far more efficient for some things. Also designing chips to come up with correct answers using the chaos and noise rather than by overriding the noise by pushing high voltage differences between 1's and 0's.
But perhaps a more sensible measurement is just to use the actual generating capacity required. 270 terawatt hours per year would be about 31 gigawatts. Consider that HydroQuebec alone produces more power than that from renewable sources, and suddenly it doesn't seem so big anymore.
I suspect that the Net saves energy. Someone using Netflix ( whose major energy consumption is from the TV, not the Internet) is not driving several miles to the theater. In the last few years miles driven per person has been decreasing, especially for the youngest drivers. Some of this comes from a bad economy, but some probably comes from substituting hanging out at Facebook, rather than a the mall.
A trip to the movies, say 20 miles, one gallon of gasoline in a typical car contains about 33KWH. By any measure this is much much greater than any concievable use for watching a flick at home. Add to that shopping online rather than driving to the mall, you have a net savings
This has been a pretty consistent trend ever since we invented electricity. Efficiency has a hard limit and our needs have none.
I believe the increases in the supply of renewable energy are on the relatively early segments of an exponential curve, whereas the growth in US requirements are presumably more linear, due in part to conservation efforts. If we are allowed another couple of decades without destroying the planet, things should look a lot better by then.
The internet has been growing exponential more or less for a couple of decades. Efficiency on the other hand can only do so well. You're not going to get exponentially improved efficiency (maybe in computing power, but not in cooling, manufacture, resource footprint of employees, etc). And efficiency is a trade off with actually doing stuff too. Try too hard to make something efficient and you will lose some degree of capability or action.
So of course, one would expect neither efficiency to be able to keep up with the growth of the internet nor for that criteria to be remotely useful to consider.
If you look at the actual article, it is about existing technologies that could allow data centers to save power. It talks about how smart standby and dynamic frequency scaling can be used to reduce the power consumption of servers.
For web sites with large server pools it would be interesting to scale the size of the load balancer pool and put excess servers into a low power standby mode.
I.e. Our server pool is 10X of what our average daily maximum network load is. This is because we get occasional traffic spikes due to [events|news stories|slashdot postings|etc] and we need to be able to scale quickly to handle the load. If we could shrink the server pool our load balancer uses to 2X and put the other 8X servers into a fast startup standby mode we would significantly reduce our daily power consumption. When the spike hit, we would put the 8X servers into full power mode and add them back to the load balancer pool. We would also put the servers into full power mode when we did software updates. This would keep all the servers current.
Approaches like this could potentially save significant power for a web site.
Another aspect of this is the old, inefficient legacy gear that could be decomissioned but isn't. Where I work there are literally dozens of old physical servers that could be virtualized. It's not being done, essentially due to mismanagement. I wouldn't be surprised if this pattern is being repeated many times over across the globe.
There should be an annual 'decomm a legacy server' day!
... assuming datacenters use 380V, just about :-) (try typing the formula directly in Google!)
230TWh / 380V / (electron charge) * (electron mass) = 12.4 tons worth of electrons
.....is almost 5.0 x 10^1000 times heavier!
ten to the thousand? Wow, that's a big number.
9x10^31 is not 1000 orders of magnitude greater than 2x10^30, it is only about 45 times bigger (assuming I haven't made another bone headed arithmetic error like everyone else in this thread trying to show off how much smarter each of us is than the last person.....)
.....is almost 5.0 x 10^1000 times heavier!
ten to the thousand? Wow, that's a big number.
9x10^31 is not 1000 orders of magnitude greater than 2x10^30, it is only about 45 times bigger (assuming I haven't made another bone headed arithmetic error like everyone else in this thread trying to show off how much smarter each of us is than the last person.....)
Don't worry, you haven't. In fact you might have the honor of being the only person left on slashdot that knows more about science and math than the average 5-year-old.
On the other hand, trying to explain math and science to an average slashdot-user is such a crazy idea that you can't possible be a logical person, so you must have guessed this value and thus you now no longer deserve the honor.
One day we will have large labour forces extracting energy for the sole purpose of feeding "the network" while governments let citizens starve and die out in the cold. And no one will find this strange. That is my darkest prediction for the future.
Actually I was just picking on the original poster for a typo.
He lost the "^" and the "-", and my brain was implicitly putting a "10^" in front of the 1031... Still a fail... but EVERYONE seems to be misreading everyone's post ... It's great!
Self proclaimed typo king, and inventor of the bear destroying coffee table (patent not pending).
Actually I was just picking on the original poster for a typo.
I knew you were razzing him about his error - but I guess you were to subtle for me (or perhaps more accurately I was too stupid for you). I thought you were just razzing him for forgetting the (-) in the exponent, whereas you were going for the difference between -31 and +1031 (and the +30 for the sun of course.)
I was feeling pretty good for a while when the AC said I was so smart, but I guess praise from an AC isn't worth much....