As someone who develops such sites for a living, I would suggest to keep things simple and just get the minimum hardware/hosting you can get by with. Time/money you spend trying to get your website scalable before it is necessary could be better spent getting and keeping paying customers. Once you have the demand, then you'll have a justification for 'moving to the cloud'.

Any kind of automatic cloud scalability is going to add substantial complexity and maintenance requirements. Unless a key selling feature of your service is scalability, I would not build for it in the beginning.

Use TortoiseSVN with beanstalkapp.com for version control, continuous integration, staging, etc. It is fantastically easy to use. You can set up an automatic deployment to a staging server on each commit, and then do manual deployments to your production server.

Another tip, you'll save on hosting costs if you go with a PHP/Linux/MySQL stack instead of Windows/ASP.NET/SQL. A single dedicated server at host like pair.com can probably host millions of hits on a carefully tuned application, and they throw in free/discounted hardware upgrades (with almost no downtime and no administrative work on your part) every few years. I could not be happier with their service.

Terrific job or white lie? The rover has lasted for 2 years, instead of 90 days. Therefore, someone made a highly inaccurate estimate at NASA. They are not supposed to make such big mistakes. Maybe they failed to realize that martian winds would clean the solar panels of dust, to keep the rover charged through the winter. Maybe NASA purposely set a conservative design lifetime as a PR move. Congress should conduct an investigation. What would happen if NASA made the same error in calculating the travel time for the human Mars mission? Oops. ;)