Barb Darrow from Gigaom published a great piece yesterday, where she both asked and attempted to answer an important question: "So… Do you really need a PaaS?" Well Barb, thanks for asking! I have a fairly strong opinion about this actually. The answer is: it depends.
More importantly, it depends on the user, the use case and the overall architecture of the application in question. Instead of talking about a specific architectural approach, we should all be talking about the needs of the users first and foremost.
Adron Hall made some similar points a week ago when he attempted to diagnose the lack of wide-spread PaaS adoption. He highlighted a significant concern expressed by many developers:
It removes key elements of what they want to control. It hides things they can't actually get to and it abstracts in ways that don't always make sense.
So have PaaS systems gotten the abstractions right? For some purposes, certainly! For other users, not necessarily (and probably not).
I'll go back to a point that I made in my last post: AWS continues to succeed because of their customer-centric approach to service design. What has this led them to create are a set of composable cloud services that span a spectrum of abstraction levels from raw VMs "as a service" up to higher level offerings like Kinesis. This customer-centricity has also allowed them to realize the value of service composition with that same variety of abstraction levels. Simply look at Beanstalk, CloudFormation and OpsWorks to see what I mean. Each of these offerings has a specific audience in mind, and has been designed specifically for that audience's architectural and abstraction expectations. This is a powerful approach.
At CumuLogic, we believe that the right solution for a cloud operator, be they public or private, is to support that same diverse array of users and architectures. After all, isn't the whole point to be supportive of your varied user communities and offer them options?
-chip
VP, Product Strategy, CumuLogic
No comments:
Post a Comment