You have a decidedly bizarre leap of logic here Ian.

The study points to gross utilization inefficiencies as a particular cause for concern, pointing out that on average, servers are typically run at 6% of their overall capacity, with data centres as a whole running at less than 60% of peak capacity.

to

So…as a message to my online networking friends out there, who never miss an opportunity to post every picture they can find to their profile, if not for the good of your friends who don’t need to know every detail about your weekends, evenings and excursion to the zoo, remember the environmental impact of your additions to the world’s data centres.

Seems like you made an easy leap of faith towards your pet peeve. The report you cite seems to argue for more efficient server architecture and utilization technologies. Yes, if every one of us put one less picture up on Facebook, that might mean Facebook needs one less server. But that also would mean Facebook has several terabytes less social interaction volume on their site, which directly impacts whatever they do to make money.

So in essence, your argument is rather like “world food prices are going up. talk less to your friends to conserve your voice and energy so you can eat less.”

The principle pursued by Alan Kay decades ago, to “waste” transistors, still holds as the fundamental axiom of computer technology.

You want a true impact analysis? Ask better questions — how much travel does virtual technology save? How much home-based work does it allow, thereby saving commutes?

I can’t resist a dig patterned after your own line: just because you can make up a quick and dirty argument doesn’t mean you should

Don’t get in the way of the Coming of the Great Big Switch with strawman arguments please. Focus on driving less.

Venkat

Venkat – thanks for the rebuttal – some very poignant remarks.

As you mention, my personal interjection re: Facebook seems misplaced and personal bias is acknowledged – maybe I just have too many eager-posting “friends” with picture folders that include “Monday, “Tuesday,” and well, you get see where it’s going…

Facebook may have been a poor choice of examples, but the point I sought to make was that our “paperless” workplaces and lives can have important environmental impacts. All of our online activity goes “somewhere” and while it may be just a drop in the bucket, much like turning off your lights when not in use – small changes, when aggregated, can have a significant impact. Like power plants, data centres are not built for full utilization all the time, but rather to manage peak demand. Peak demand becomes “peak” with the aggregation of many individual, isolated activities (individual online activity).

I agree that technology has (and will continually) facilitate important reductions in travel and commutes and do not claim otherwise. You are bang on there.

Kudos on the dig

Ian makes a good point – the paperless office is not without its own environmantal impacts.

To put into persepective the tradeoff, consider the fact that even back in 2006 Google spent more on powering their data center than the Wall Street Journal did on paper. In 2005, the WSJ reportedly spent $115 million a year on newsprint for all its publications. Although Google keeps the details of its computing infrastructure a highly guarded secret, most estimates from from the time pegged the company’s facilities at approximately 500,000 fully-redundant servers (and growing). With the average high-efficiency server using approximately 300 watts of power, and an average commercial energy cost of $0.09/kWh, Google spent as much as $118 million a year, just to power their servers (not including cooling, switching, and conversion, which would easily double this figure). This is probably part of why Google invested in a 1,900 kW solar panel project back in the Fall of 2006 (http://googleblog.blogspot.com/2006/10/corporate-solar-is-coming.html).

In terms of data center utilization and efficiency, there are certainly a lot of thinks enterprises can do, and are doing, including: server virtualization (to improve capacity utilization), airside economizers (to user outside air to cool the data center), and proper data management (to prioritize data so that not everything gets duplicated for backup). In fact, most large data center companies (like Google, Facebook, and Microsoft) have this stuff figured our because its impossible to ignore the cost argument. The real offenders are likely small- and mid-sized shops that individually have little incentive to make big investments in improving efficiency, but collectively account for significant under-utilization and waste.

This means that there are additional savings to come from user initiatives such as deleting old messages and resisting the urge to post photo diaries of Rover’s first trip to vet or the visual progression of intoxication from every Saturday night. These small steps don’t detract from the value of social networks but do help the environment. Ignoring these incremental efforts is similar to the “it’s just one pop can” argument against recycling back in the ’80s. As Ian mentions, every little bit counts.

See Bill St. Arnaud’s recent presentation at the BC Net conference on this topic. He is looking into researching greener server technologies and locating them in northern rural places in order to power them with green energy and to reduce cooling costs:

https://wiki.bc.net/atl-conf/display/BCNETPUBLIC/The+%27Greening%27+of+British+Columbia%27s+Research+Networks

Hmm… I am going to stick to my guns here because people insist on making value judgments on what is ‘frivolous’ and what is ‘serious’ when it comes to using computing. Naumi’s “photo diaries of Rover’s first trip to vet or the visual progression of intoxication from every Saturday night” examples are a case in point. I’ll assume that those of you making this argument somehow consider it as a given that ACH transaction data, say, is self-evidently more “important” than Rover’s first trip to the vet.

These are not sociologically frivolous in their ultimate impact on our world, even though the original intent may be frivolous. These help create a broader form of what people are starting to call ambient presence and belong in the bucket of soft goods allow social computing to create a true virtual geography and virtual social capital. I’ve written a series of pieces exploring this, and while I can’t claim the sociological questions are all resolved, it is a deep mistake to presumptively assume that massive, large-scale social behaviors are irresponsible frivolity. They just might be fulfilling some function critical to the world’s sanity.

I am not arguing we shouldn’t work towards energy efficiency in servers. We just shouldn’t pick our battles based on judgments passed on others’ social behaviors. Critiquing photo-sharing behaviors is ultimately as naive as saying, for example, that Christmas is a worldwide waste of resources, ignoring the social role of that holiday. If you are not prepared to do the deep due diligence required to analyze why a social behavior exists, it is irresponsible to suggest it should be reduced.

The analogy to soda cans is misplaced. These are two very different problems in environmental impact analysis.

Ian — you’re a sport

Venkat

I will take a different approach to this issue than the one about judging the frivolity of the social content.

Instead, I call BS on the McKinsey study, not for having faulty research, but for drawing the wrong conclusions.

Saying that server farms will produce more carbon emissions than the airline industry is based on the continuation of today’s technology, with the increasing usage we see. However, as we’ve seen over the last 30 years, technology improves exponentially, and we have a lot of room to go to continue improving the efficiency of servers, including solid state drives which are becoming common place and more efficient processors.

In addition to the technology, this study assumes that we will be powering server farms with the same fossil fuel driven electricity that we’re using today. Jets for the foreseeable future will only be able to fly with carbon emitting jet fuel. Servers can be powered by anything, including the improving solar cells that are such in vogue by the many eco conscious technology companies. So even if the technology doesn’t improve, I still can’t see server farms producing as much CARBON emissions as the airlines.

The logic just doesn’t hold.

[...] 22nd, 2008, 06:24pm In May Ian had an interesting post about the environmental impact of data centers – the power hungry workhorses of the internet. As [...]