When 1st Life Meets 2nd Life: The 1685 Pound Avatar and the 99 Ton Acre


A Second Life avatar produces 1,685 pounds of CO2? And an acre of real estate in Second Life produces 99 tons of CO2? What gives?

One of my pet research projects has been to find ways to establish effective, playful bridges between 1st life and 2nd life. What is 1st Life and what is 2nd Life? By 2nd life, I’m not referring to the virtual world run by Linden Labs. For me, “2nd life” is an evocative metaphor that counterpoints the normal, human, physical, material world, which I refer to as 1st life. I’ll grant that the language lacks precision, but I’m relying a bit on my own assumptions, which I think are fairly well-shared, as to what counts as a digital, networked, social environment. 1st life is then the non-networked, non-digital social environment.

The presentation — titled When 1st Life Meets 2nd Life I gave this week at Lift07 started with a reminder as to the material basis of 2nd life. There is “stuff” that undergirds digital networks — indeed every digital bit owes its life to some sort of material. Atoms compose digital data. There’s that stuff that we’re never really aware of unless we spend time working in the data center facilities where all of the Internets take physical form. 2nd life and its digital networks are made of heavy material — copper cable, steel racks to hold servers, rubber or plastic insulated power cabling, cooling systems, human labor, billions and billions of integrated circuits and the effluvia of the toxic chemicals expelled during their production, shipping and decomposition, etc. Our participation in this materiality probably ends at about the time when we discard the cardboard box and styrofoam packing of the shipping material in which our new computers arrive.

But this is more than the William Mitchell bits & atoms thesis. It’s not just the equivalence, but the precise nature of that material — what kind of stuff are we talking about? It’s not just the composition of digital bits, but what physical material, in the use of digital bits, of digital networks, of our PC’s, web and game servers, is produced.

Perhaps the most unsettling material characteristic of our 2nd lives is considering the resources necessary to maintain them. Whether emailing, googling, blogging, uploading videos, downloading music — everything — this owes a measurable and material debt to first life. When one computes the amount of power — normal 1st life electricity — that is consumed to maintain our 2nd lives it becomes clear what that debt is. Or when one computes how many tons of CO2 emissions result from the production of that electricity, assuming the majority of our power comes from plants that burn something that produces CO2 as a consequence of producing electricity.

I started running these numbers after reading a very interesting and thought-provoking discussion on Nicholas Carr’s blog where he computed these figures based on some public figures he found pertaining to Linden Lab’s Second Life environment. One could run the same numbers for any other digitally networked activity, like emailing or web surfing or whatever.

What’s particularly appealing about choosing an online world like Second Life as my example is that it’s underlying metaphor is 1st life. Email would be another good example, as postal mail requires energy that exhausts CO2 in its processing and delivery. But Second Life has more PR these days, so I’ll use that as an example.

Second Life itself captures many of the important characteristics of 1st life and uses that to convey a sense of familiarity for the users. There is property in Second Life, waterways, buildings, etc. It’s a 3D virtual world that is largely modeled on commonly held assumptions about what counts as 1st life. What it doesn’t convey to its users is any kind of Second Life representation of the ecological cost of that Second Life world, which would be very cool — Second Life CO2 emissions, for instance, to correspond to equivalent estimates about how much CO2 is emitted in 1st life.

I was shocked at the numbers on Carr’s blog, so I computed them myself to check the math. I revised some of his assumptions, so my figures are significantly more conservative than his. (I’d even go so far as to say that my figures are unrealistically low, because a more rigorous analysis would include estimates about the power consumption of the ancillary network devices between the user’s computer and the data center.) I also culled from comments in Carr’s blog post to refine some of the assumptions, especially the remarks from Second Life employees who have direct access to some of the power consumption figures.

There are a few additional assumptions I’ve made, mostly pertaining to what I think is a more realistic assumption as to how much power a typical home PC uses, and how often one might actually play Second Life.

Carr assumes that a home PC consumes 120 watts, which I think is much too low — I assume 300 watts, based on looking at the technical specifications of a mid-range Dell computer, and I also compute the power consumption of an LCD display. I also don’t assume, as Carr does, that someone playing Second Life is playing 24 hours a day — I assume, averaged over a year, they will play eight hours per day. Some days they won’t play, others they may invest 12 hours. I think 8 hours is a fair assumption.

I’ve also used the assumption that 1.35 pounds of CO2 is emitted per kWH of electricity produced.

A Linden Labs employee measured the power consumption of their servers and came up with the figure of 175 watts (energy per hour) with the server running at full-tilt. I assume that the servers basically run at full-tilt 24 hours a day, seven days a week, and that servers of this sort make demands on the data center for cooling, power distribution, ancillary resources like lighting, operations center energy costs, keeping the candy and pop machine running in the break room, etc., at an equivalent of 50% of their nominal energy use. So, a 175 watt server actually needs 175+87.5 watts of energy to function in a data center.

I came up with the following figures (the calculus is further below):

Power Consumption Per Avatar Per Year (Second Life Servers): 153 kWH
Power Consumption Per Avatar Per Year (Home User’s PC): 1,095 kWH
Total Power Consumption Per Avatar Per Year: 1,248 kWH

CO2 Emissions Per Avatar Per Year (Second Life Servers): 207 lbs (94 kilos)
CO2 Emissions Per Avatar Per Year (Home User’s PC): 1,478 lbs (670 kilos)
Total CO2 Emissions Per Avatar Per Year: 1,685 lbs (764 kilos)

Second Life is composed of regions that have a correspondence to normal 1st life acres. I’ve learned that there are 16 acres per region, and there are 4 regions per server, so there are a total of 64 Second Life acres per server. That means the power consumption per Second Life acre is 4.10 kWH, or 147,168 kWH per Second Life acre per 1st life year. And that means that 23 pounds of CO2 is produced and exhausted into the 1st life atmosphere per Second Life acre per hour, or 198,677 pounds (90,118 kilos) per Second Life acre per year.

Some equivalence for perspective:

In 2003, the per-capita power consumption in the United States: 13,242 kWH
In 2001, the per-capita power consumption in Iceland: 26,947 kWH
In 2001, the per-capita power consumption in Keyna: 118 kWH
(World Resource Institute, EarthTrends — http://earthtrends.wri.org)

Every year, every Avatar in Second Life produces CO2 emissions equivalent to a typical, honking, bloated, arrogant SUV driving 1,293 miles, based on the assumption that this kind of SUV generates 1 lbs of CO2 per mile.

If serving this page took the CPU on my modest web server .1 second to serve to you, it probably consumed .004 watt (assuming my clunker consumes 100 watts + 50 watts for overhead)..it works out to about 6 micrograms of CO2, not counting whatever your PC contributed to the production of CO2. Okay, I’m getting carried away, but you get the idea.

Why do I blog this?I find this kind of analysis fascinating and revealing. It is the kind of bridge between 1st life and 2nd life I am trying to build, where the semantic link between what goes on in our 2nd life worlds is made plain in its correspondence to 1st life. 2nd life activities are not the clean, sustainable, whole-earth friendly activities the Bay Area, Web 2.0 crew may think they are. Despite the important evolution of human social formations that have arisen, the messiness of the 1st life remains. Maybe there should be a little eco-meter on the dashboard of Second Life, World of Warcraft and, whatever — YouTube and my blog. I’d be interested in computing the same figures for World of Warcraft. I suspect they’re probably equivalent, although I’d probably bump up the average hours of play per WoW character quite a bit. I would need to know the distribution of simultaneous characters per server, or number of servers per instance, as well as some sense as to how much power is consumed by whatever server they may use. It’s about provoking some thinking about the material contingencies of our online activities. I won’t really quibble about the numbers. Someone’s going to want me to adjust something this way or that — the accuracy of the figure will be forever elusive, so I’m not interested in debating that, or tweaking some of the numbers. The point is — there is a debt paid for our online lives and we rarely think about it. How can we start to introduce the material aspects of this activity more directly? That is my goal here.

The Calculus

At The Data Center
Server Power Consumption (Energy Per Hour): 175 watts
Percept Power Consumption for Cooling and Ancillary Network Operations Center Needs: 50%
Ancillary Power Consumption Per Server (Energy Per Hour): 87.5 watts
Total Server Instance Power Consumption (Energy Per Hour): 262.5

Active Servers for Second Life: 1,000
Active Avatars in Second Life Per Day: 15,000
Average Active Avatars Per Server in Second Life Per Day: 15
Power Consumption Per Avatar Per Hour (Server Only): 262.5/15=17.5
Power Consumption Per Avatar Per Day (Server Only): 17.5*24=420
Power Consumption Per Avatar Per Year (kilowatt hours): 420*365/1000=153.3
CO2 Emissions Per Avatar Per Year (lbs): 153.3 (kilowatt hours)*1.35 (CO2 pounds per kilowatt hour)

Acres Per Region: 16
Regions Per Server: 4
Acres Per Server: 64
Power Consumption Per Acre: 262.5/64=4.10 kWH
Power Consumption Per Acre Per Year: 4.10*24*365=35,916kWH

CO2 Emissions Per Acre Per Hour: 1.35 (CO2 pounds per kilowatt hour) * 4.10 kWH = 5.535 lbs
CO2 Emissions Per Acre Per Year: 5.535 *24*365=48476.6 lbs

Average PC Power Consumption: 300 watts
Average PC LCD (17″) Power Consumption: 75 watts

At Home
PC Use (Hours/Day): 8
Power Consumption Per PC Per Day: 3 kWH
Power Consumption Per Avatar Per Day: 3 kWH (One PC At Home Per Avatar)
Power Consumption Per Avatar Per Year: 1,095 kWH

CO2 Emissions Per Avatar Per Day (Home PC @ 8 hours/day): 1.35 CO2 Pounds Per kWH * 3kWH = 4.05 lbs
CO2 Emissions Per Avatar Per Year (Home PC @ 8 hours/day): 4.05 * 365 = 1,478 lbs

Power Consumption Per Avatar Per Day (Home + Server): 20.5 lbs
Power Consumption Per Avatar Per Year (Home + Server): 1,248.3 lbs

CO2 Emissions Per Avatar Per Day (Home + Server): 27.68 lbs
CO2 Emissions Per Avatar Per Year: 1,685 lbs

SUV CO2 Emissions Per Mile: 1 lbs
Equivalent SUV Miles Per Avatar Per Year: 1,685

Additional Links
Data Centers’ Growing Power Demands: A new report quantifies the electricity consumption of servers, revealing a startling trend.

Demand grows, but data centers don’t hog power: Net uses barely over 1 percent of U.S. electricity, study says

Parenthetically, I’ve been asked why I pluralize Internets isn’t there just one? I guess my speculation is that we are heading toward a kind of Balkanization of digital networks, thanks to the deleterious effects of net neutrality. It wouldn’t surprise me if there were tiers of networks, with levels of privilege offered based on one’s ability to pay for better or faster or thicker bandwidth.

Technorati Tags: , , ,

5 thoughts on “When 1st Life Meets 2nd Life: The 1685 Pound Avatar and the 99 Ton Acre”

  1. Hi Julian,

    I’ve been following the CO2 per SL Avatar debate with interest across the web and have come across many differing opinions on the topic, not least of which from Nicholas Carr’s original blog post on the subject.

    It’s great to see you picking this up and applying some of the findings from within the Carr comments to come up with your new figures and findings.

    I had one query on your figures though which I couldn’t understand. If you could help, that would be much appreciated. please forgive me if I have misunderstood/misinterpretted anything!

    You say the power consumption of one SL server is 262.5 watts per hour. Also one SL server hosts 64 acres of land so you calculate that watts per acre is 262.5 watts x 64 (totalling 16.8kwh per acre per hour = 147,168kwh per acre per year). My query is that, if the 64 acres share a server using a total of 262.5watts per hour, shouldn’t the power per acre be the 262.5 watts divided by the 64 acres = 0.004kwh per acre per hour = 35.9kwh per acre per year?

    As I say, please forgive me if I’ve misunderstood but i’d be grateful if you could explain / correct me.


  2. Hi Steve,

    I think you may be correct. I multiplied when I should have divided. I’ve corrected the derivation.



Comments are closed.