Winpdb - A Platform Independent Python Debugger Elastic Plastic
Can someone enlighten me to the fundamental difference between Amazon’s EC2 service and a standard dedicated hosting service?
Today I chatted with a support person from Rackspace “cloud” servers and tried to figure out if there is any technical difference at all between their service and Slicehost’s VPS hosting. “No” she admitted, “They are identical”. They even use the same server specs.
Dynamics.
In standard hosting service you buy 1 (or 2, or 10) hosts and keep them. From time to time you can add next one, but in general it is fixed. In EC2 you can boot extra machine for 6 hours, then shut it down, and be charged for those 6 hours. Or have more systems running afternoon and less at night.
Of course the true point is in the things which just start to appear - having instances boot/shutdown automatically according to the load.
If you buy single EC2 and keep it running, there is surely no benefit comparing to normal hosting (and some hassle). If you need 100 machines at peak, you can find this offer interesting.
SmugMug owner blogged some interesting posts about the way they use Amazon (http://blogs.smugmug.com/don/)
You can find a 3TB 4 core xeon (referred to as extra big plastic by Amazon) for 150 - 500 USD per month.
Even if you end up using half the bandwidth and your servers sit idle half the time it would still be a better deal than Amazon’s.
So what’s the point in all this “flexibility” and pay for what you use billing if it ends up more expensive anyway?
I’m running a big index task right now. It’s going to take about 2 days on a 4 core AMD box maybe comparable to that “big plastic”. OK, I can rent a box like that for (say) 300 dollars a month, which comes out to $10 a day. But if I can rent it by the day for $20, that can be more attractive if I only want it for 2 days instead of 1 month. It gets even more interesting if I can rent as many of those machines as I like for $2/hour, because then for $100, I can run my task on 50 machines and be finished in 1 hour instead of 2 days. THAT is the attraction of cloud computing. The minimum rental interval is hours rather than months.
Indeed, that is the most defining practical distinction of “cloud” computing - billing of actual resource consumption.
But at the same time this is also a ridiculous distinction. Is this the essence of this technological revolution? usage-billing?
Take a listen to this presentation (http://bit.ly/8rwGUH) it is both humorous and informative. I also agree that the concept of a “utility” service is not new.
The benefit I get from using EC2 is that I can get a machine now and only for 2 hours if I need, not have to wait hours/days to get an agreement in place with a minimum of a month.
For dedicated machines I do find EC2 relatively expensive, I like hybrid services like Servepath/Gogrid and others offer.
Thanks for asking this question. It is something I’ve been wondering about.