The article by Leavitt , on January Computer magazine, is a nice, and in my view impartial, discussion on what cloud has and can bring in to the table. It points to few problems that are still at large.
Also the blog post  by James Governor, argues that hardware as the service model of Amazon is the way to go. I agree that nice simple model of AWS simpler, because it enables users to port with relatively less changes, and of course "simpler is better". However, the higher level of abstractions like App Engine could provide more features like failover, or auto scaling (I mean auto scaling the application, not just increasing machines). Among questions are how general will those features, how much changes do they need to exiting systems, can they solve associated hard problems?
 Neal Leavitt, "Is Cloud Computing Really Ready for Prime Time?," Computer, vol. 42, no. 1, pp. 15-20, January, 2009.
 Amazon Web Services: an instance of weakness as strength