Kief Morris
Jul 24, 2017 · 2 min read

I agree that building with PaaS and Serverless saves massive amounts of time and money over building your own automated infrastructure on top of IaaS cloud. I also see IT teams putting up roadblocks, FUD, etc. to keep things the way they’re used to.

But there’s another big blocker, which is application architecture. Using PaaS, and even more with serverless, requires your applications to be built from the ground up for that model. For a green field project, or a major rebuild, the target architecture should be these. But most organizations moving to cloud have a fair bit of existing code, which is non-trivial to rebuild for a radically new architecture.

So IaaS offers a way to migrate to cloud for these people. It has costs — automating a cloud-based infrastructure is non-trivial. And ideally it’s a stepping stone, that allows new code to be written for a more modern architecture. Your experience may vary, based on the types of clients you work with — in my experience, medium to large, more established businesses face a multi-year journey to completely rebuild their systems to be truly cloud native. So IaaS is a valid, pragmatic choice for quite a few people.

I’d also suggest that positioning this as Google vs. Amazon is a red herring. Google has an IaaS platform (GCE), which is just as expensive to automate as AWS EC2. And AWS does offer serverless (Lambda) and containerized services (ECS), so you can build a completely cloud-native system on AWS, with all of the cost savings offered by GKE or AppEngine.

I don’t disagree with your article, BTW, just pointing out there are more angles to the story.

    Kief Morris

    Written by

    Cloud Practice Lead at ThoughtWorks. Author of “Infrastructure as Code”, from O’Reilly.