Failure to Launch

When I attended the Cloud Camp last November I had run a session called "Failure to Launch" which was members of the audience talking about their early experiences with Cloud Computing. I was looking for some lessons learned and possibly some unique risk factors associated with Cloud Computing. What I heard was that projects in the cloud are influenced by factors common to most other emerging technology projects. The best example was given by Uri Budnik of RightScale. He had to keep the customer anonymous but did share the following:

Name of Project - Planned Major News Event for major news media 

Project Dates - Project began three weeks before hard news deadline

Technologies used - AWS EC2, RightScale, HAProxy, DNS round robin, Mongrel, My

What happend? - The system was unacceptably slow in early versions and could not be improved. Some of the content would not load. The customer introduced a last minute architectural change the morning of the event that required rollback in order to launch. 

Lessons Learned - Need more through testing.

I heard some similar profiles from other participants about classic software engineering problems...  scope creep, lack of testing, lack of communication with stakeholders, unrealistic schedule expectations.