post

Is SaaS Green?

Back of envelope calculations

Is a hosted app more Green than hosting your own? Is Software as a Service (SaaS) more environmentally than the more traditional models?

I contacted a number of SaaS vendors (who, admittedly, may have a vested interest in this!) but the answers were a resounding yes.

Chris Yeh, for example, from PBWiki did some quick back of the envelop calculations and replied with:

PBwiki hosts 500,000+ wikis on a total of 20 servers
If a server consumes 200 watts of power, that’s 1.75 megawatt hours/server/year (200*24*365)
According to Sun’s Dave Douglas (http://www.roughtype.com/archives/2006/12/avatars_consume.php), that’s the equivalent of 1.17 tons of carbon dioxide per year, or driving an SUV 2,300 miles
That means PBwiki could be saving the world up to 585,000 tons of carbon dioxide per year, or the equivalent of driving an SUV around the world 50,000 times!

Now, obviously not all 500,000 hosted PBWiki’s are replacing an individual server but say 100 PBWikis replaces one server, or 1,000. That’s still somewhere between 600 and 6,000 tons of CO2 PBWiki are saving the planet per annum.

Anyone got harder numbers than that around the energy efficiency of SaaS?

Comments

  1. says

    Seems like an awfully complicated one to tackle. Systems like Amazon EC2 likely keep servers hot so they can be quickly made available to users. So there could be many servers doing nothing useful but consuming electricity and generating heat. On the other hand that kind of system can probably be better optimised than if the servers were spread amongst many companies.

    Also a small company could power down servers over weekends when they are not used. But if they were using PbWiki the server wouldn’t be powered down (shared or not.)

    Awfully complicated I’d say.

  2. says

    complicated yes but you have to start somewhere!

    any other replies yes Tom? we should perhaps make it a bit more formal. wondering how to tabulate

  3. Asa says

    This is a big issue for universities, too. Most of the energy growth at Caltech over the last few years, and the projected growth going forward, is high performance computing. We most definitely want to know if an off-site virtualized solution (ie cloud computing) is better than our own server rooms (highly likely that it is), and most importantly, how much better? We need to be able to go a faculty member and offer them the chance to trade out to an off-site solution, and know how much energy this will save.

  4. Ludovic says

    You need to add the utilisation rates to get a complete picture… Intel is typically 10-40%, mainfraimes are 70-85%…

  5. says

    Tom,

    Just to revisit this issue from last year…we’re now up to 900,000 wikis hosted on a total of 67 servers. Slightly less efficient on a per-server basis, but still pretty darn good.

  6. says

    I am particularly interested in Cloud computing and I go with SAAS as Green, You never need to make anything yours and having a centralised service can minimise the power consumed when performed seperately.

  7. says

    I read somewhere last yjear that Microsoft had some servers that were very impressive in this department and I know that as time goes on carbon footprints will grow smaller and smaller. Thanks for the post.

Trackbacks