This article from RW was interesting. I remember I did a double-take when hearing that Zynga was leaving AWS but there was general agreement at the time that you could save money by hosting it yourself (as well as some flexibility) if your environment aligned to that model, but it wasn't clear what the formula was. To me it's clear that AWS has an advantage when the use case is not static environments. When your needs change, you aren't dragging the ball and chain of past decision around. This is especially true for big-data efforts and emerging businesses.
But what is static these days? Businesses come and go. Product lines change. Customers needs change. State of the art in compute and architecture changes. Its' difficult to find the use case where change isn't a persistent march. Where change is slower, SaaS product prevail by taking a mature service and cost-optimizing it for their customers (ala CRM, email, etc.). Here is a small quote from the article.
- As the company tried to make the shift to mobile, however, its infrastructure did it no favors. As Amazon Web Services data science chief Matt Wood told me in an interview:
Those that go out and buy expensive infrastructure find that the problem scope and domain shift really quickly. By the time they get around to answering the original question, the business has moved on.
- While Wood was speaking of solving tough data science problems, the same principles apply to IT infrastructure, generally. When you build out a data center to solve particular problems, you're stuck with infrastructure that may not suit itself to new business challenges that arise.
There are a lot of advantages of cloud computing, but this is might be a really good example of the agility advantage.
-- Chris Claborne