September 28, 2009
I’m a bit of a curmudgeon – recently I was critical over lily-gilding in the cloud computing space – in that instance it was a case of someone holding out a cloud deployment as something way more than it was. This time it’s the turn of greenwash to come under the spotlight. Recently Chris Thorman wrote a detailed post looking at whether SaaS or on premises software is more “green”. The post generated a number of comments, mainly from other cloud computing evangelists who used it as yet another point of justification. In the spirit of CloudAve encouraging diverse opinions, I thought I’d chime in with my thoughts.
Chris’ analysis was made using the assumptions detailed in the following diagram;
His example was based on a medical practice with four physicians, initially using traditional on-premises software and then moving to SaaS.
In the former case, he’s details energy use for a desktop per physician plus a local server. Already I’m kind of dubious – a number of organizations I deal with of the size used in this example run a desktop machine as a server. In this case that would mean four desktops, one of which doubled as a print/file server. In that case you’d have a (roughly) 2600 KW/yr desktop/server plus three 600 KW/yr desktops for a total of around 1200KW per user. A commenter on the post also pointed out that;
You indicate that a Dell computer draws 300W of power each hour? Have you ever measured that? An inefficient computer will draw less than 100W, and a modern computer will draw closer to 50W.
Now things get interesting – perhaps in an attempt to fend of the naysayers who would say that medical records can’t be run in the clouds for privacy reasons (I disagree with that contention but no matter), Chris has modeled his analysis on running twin dedicated servers within a Rackspace data centre. He’s also made the assumption (wrong in my view) that by merely moving medical records to SaaS, all of a sudden our physicians will move from fully featured desktops to netbooks. I have an issue with that on a couple of levels;
- Most people that I know who have spent time working on a netbook don’t really believe they’re viable for day in/day out work. Our physicians would still require “real computers”
- A desktop to netbook comparison isn’t apples with apples – more fair would have been comparing a full featured notebook with a netbook
- As a commenter pointed out “…there’s nothing stopping you from running a web-based application from an onsite-server, so you’d be able to switch to netbooks if you wanted to. Saying that netbooks can only be used with SaaS services is just plain incompetent”
Chris does go on to point out that using multi tenancy, the server energy consumption component can be hugely diminished (in his example down to 1% of the total server energy requirements or 130KW/annum).
Chris finished up, declaring from the pulpit that;
Using SaaS, our four physician practice now only consumes 611.4 KW per year running their EMR software. That’s 152.85 KW per year, per physician.
That’s an 93% reduction in overall energy consumption for a four physician practice using SaaS EMR software over on-premise software!
The problem I have with the article is that it’s another example of the power of statistics to tell whatever story the write wished. I’m an evangelist for cloud computing and, yes, it is correct to say that it is generally more efficient to run a large efficient data centre than a small private server set up – but much of Chris’ argument is fallacious – and it’s by traipsing out fallacies that we, as industry evangelists, get caught out.