A co-worker brought up the subject of how much electricity Yahoo and Google use to run their server farms. His claim was that both of these companies had intentionally located some of their major server farms close to power plants! Maybe this is a slight exaggeration, but his main point was valid: to run something on the scale of Google or even Yahoo requires a huge array of servers that, on average, run at a small percent of their total CPU capacity.

This got me to thinking, how much electricity is being used just so that you can look at this web page? It sits on a server (or servers) that are idle most of the time. In addition, there is a lot of telecom equipment carrying the data between you and the servers, and this equipment must be on 24/7, regardless of how much data is being shoved through it.

Taking this even further, the BBC and other large international broadcasters once used millions of watts of electricity to run several shortwave transmitters simultaneously. Now, much international broadcasting is done via Internet streaming. Is this more efficient, in terms of energy used per listener, than shortwave radio was? I wonder if anybody has ever tried to do an analysis on this.

Interesting topic.

I do know there is a lot of research and development going into cluster computing. GFlops / Kilowatt is a measure of growing importance.

We are just now really starting multi-processing. Software is in a crude state, as is our understanding of how problems might be solved.

One biggie is peak compute. It's true yahoo or google has a fairly high number of wasted cycles. However, they also have need, on each node, for serious performance at times. Working through that issue will reduce power consumption.

Light based computing may well do this as well.

In my mind, this is a biggie. Being able to get along well enough to take better advantage of distributed computing might help this issue somewhat. Coding in ways that perhaps consume more working space, but far less peak compute might help too.

One positive is the utility we get from the energy use! I'll take the Internet over SW anyday.

More distributed solutions may well be better able to take advantage of sustainable power sources more of the time as well. Redundancy, combined with higher bandwidth networks could also move the load to where the power is.

...or just do what Google wants to do and turn the moon into a huge data center, powered by solar, cooled by the always cold conditions there! We trade some latency for a serious energy savings, and some insurance too. Should we all go nuts and blow the place up, the survivors will eventually get to leverage all we learned!

I suspect we've just begun to broach this topic.

The subject that led my co-worker into the discussion of server farms that use lots of electricity was hardware virtualization, which might be along the lines of what you were thinking with multi-processing.

The concept that he described could be illustrated like this: Imagine that you have three servers, server A, server B, and server C. On average, each is running well below 33% of its capacity. However, three servers are being used either because the software won't all run on the same OS/hardware or because it is very important that even if a program manages to crash one of the servers, the software on the other servers contine to run without interruption.

With virtualization, all of the software could be put on a single server, cutting the power consumption. Of course the two drawbacks to this approach are, one, that you lose the peak computing power capabilities that existed when running three servers and two, that if your server goes down due to a hardware malfunction, you lose big time.

I did a Google search on "environmental impact of the Internet," and I saw some interesting results. There are people who are trying to perform this type of analysis, but it gets to be a Herculean if the question is a specific one like, "how do the energy usage and carbon emissions from printing a newspaper compare to reading that newspaper online?" I think that we might never have a satisfactory answer to specific questions like these. However, I think that a more plausible question is, how will people 50 years from now look at our energy usage today, in light of all of the new server farms and telecomm equipment that went into service in the late 1990s and beyond? And, what about the fact that as workplace personal computers became ubiquitous during that time frame, people generally got into the habit of leaving them on all the time?

Dont forget, the internet, using e-mail cuts way down on the amount of physical mail having to be delivered. Using video conferencing cuts down on travel, By using on line documentation cuts down the amount of owners manuals etc being printed.

Paper use saw a dramatic increase with the advent of email.

http://earthtrends.wri.org/features/view_feature.php?theme=6&fid=19

Paper use has increased some, but not dramatically, and it would be a streatch to blame it on e-mail.
At the same time e-mail was becoming widly used, there was a hugh increase in the availability of personal printers, and high speed printers for corporate use.
The cost of direct mail marketing also dropped a lot, due to the use of high speed printers and advance word processing/mailer software that let marketers send more "personalized" mailings.
Its a myth of environmentalists that e-mail has greatly increased paper usage.

Printers saw a dramatic increase with the invention of printers then.

In my own experience, networked computing has lowered my energy footprint considerably.

I travel perhaps a fifth of what I used to. Paper use has finally gone down too. High resolution displays and reasonable portable computers are finally able to serve in place of paper most of the time.

Alfredo: Exactly!

Multi-processing machines are ideal for virtualization. It's a brute force approach, but it does take good advantage of existing single threaded code.

I'm currently setting up a 64Bit Linux machine that will host as many as 8 virtual machines, all running at the same time. Essentially, I can get rid of an entire bunker of support machines in one swoop. Companies are doing the same things and it's all really ramping up right now.

When we get better at writing multi-processor friendly software, we will be able to make use of many lower speed chips instead of a coupla really fast ones. The power implications, for many devices are just huge.

Imagine a laptop running all day. This could happen with multi-processing. Better, imagine a laptop that taps other computing resources nearby, via network. That's all multi-processing type stuff and it's damn cool.

Or, a bunch of laptops that run faster when in close proximity...

Which is it, a "streatch" or a myth?

Point is that the "paperless office" or whatever you care to call it actually uses more paper than the old days.

Calling it a myth (and taking it as an excuse to take a potshot at your political enemies) is splitting hairs.

I'd have my doubts on the online documentation bit as well- my impression is that people tend to print any doc they want to use extensively... then they'll print it again next time they need it!

I agree with this, for the majority of people right now.

However, we have great display tech coming down the line, that's gonna change that. Electronic paper, type displays are gonna be found quite useful for a lot of people. Legal concerns will, rightfully, mandate the continued use of paper, but that's actually not a bad thing.

Getting rid of casual uses of it will have a pretty serious impact, IMHO.

I will still print docs, but I've found this less and less necessary.

It takes some different habits and ways of working to use electronic docs.

I said e-mail cant be poven to have increased paper usage, because it was part of the technolagy revolution, there are many other aspects o the growth in technology that lead to higher paper consumption.

Have you ever heard of another wonderful "poven" tool of the "technolagy revolution"?

It's called spellcheck.

"Point is that the 'paperless office' or whatever you care to call it actually uses more paper than the old days........my impression is that people tend to print any doc they want to use extensively... then they'll print it again next time they need it!" (qtd. from Tadc)

"Electronic paper, type displays are gonna be found quite useful for a lot of people. Legal concerns will, rightfully, mandate the continued use of paper, but that's actually not a bad thing..........Getting rid of casual uses of it will have a pretty serious impact, IMHO." (qtd. from Missing)

One fact I have noticed constantly being overlooked by the "Paperless Office" camp:

Computers are NOT at all 100% reliable and in the long run are a *very unreliable* method of archiving important documentation!

Yes, you should make backups of your system periodically so no data are lost. (Likely many of you here do!) Unfortuantely it is a harsh reality even backups have their shortcomings--tapes could get erased accidentally, DVDs/CDs could become corrupted, heck even the other hard drive in your RAID array could fail unexpectedly!

And there is always the possibility of *computer viruses or malware*!!

Don't get me wrong, I do think it is a very cool concept, computers/networks are convenient and can be *way* easier to manage than a room full of vertical/hanging files.......but it unfortunately can never replace the hard copy in terms of reliability and longevity.

If it is something important ALWAYS print it out as a backup or reference! (That is what I do.....a copy of my work on the screen--and disk--and a copy bound in my notebook.) Computers can crash and networks can bomb unexpectedly at any time often without any warning--paper usually doesn't.

Just a reminder...............

OK.

I use more paper because of email. So does our business.

There.