An Open Letter To Embrace AWS And What It Says About OpenStack’s Self-Serving Vendors

Cloudscaling CTO Randy Bias wrote an open letter to OpenStack today. In it he outlines why the open cloud effort will only win if it accepts Amazon Web Services (AWS) and creates a compatible API.

He argues that AWS is the defacto leader. The solution: OpenStack should stop trying to build out its own differentiated APIs and accept the reality that AWS is the winner in the public cloud. If it does that then OpenStack can win in the “hybrid” cloud where the AWS-style public cloud meets the modern data center. This is where OpenStack can soar — helping customers adapt by offering a cloud operating system that has its own elasticity but without the scale of a massive service for tens of thousands of customers.

In particular Bias calls out Rackspace for trying to make its own API as the standard for OpenStack. He details the history of how OpenStack has used APIs to favor the Rackspace public cloud. Bias notes that Rackspace did all of this to serve itself. It used OpenStack to try to differentiate its own service.

That’s certainly true. There is no doubt that Rackspace saw an opportunity to turn itself into something much bigger by creating the OpenStack organization. The company struggled with the new way of the cloud. It had to turn from being a hosting company to a software developer. Did Rackspace not quite get it? It now seems that way. By trying to play the role of leader it also sought to seek control and try to force its own cloud by calling its own API the “native,” one for OpenStack.

To its credit, Rackspace also rang the rallying bell for the first major open cloud initiative that today has more than 250 vendors participating and thousands of developers who have written more than 1.2 million lines of code. IBM, Red Hat and HP all joined OpenStack, and it gave Bias a window into a new market for Cloudscaling, which provides its own system for building out cloud infrastructures.

But Bias has his own reasons for putting such emphasis on AWS. It serves his own interests. His company has placed its own bets on AWS and Google Compute Engine. APIs to overlay AWS and OpenStack would certainly do a lot to help the cause of his young company. For more on this angle, I’d recommend reading cloud commentator Ben Kepes post for his perspectives on the matter.

It’s a curious set of circumstances. The powerful and not so powerful vendors in OpenStack all face considerable market pressures, which are exacerbated by AWS and its unquestionable innovation. In the three years since OpenStack was founded, AWS has just dominated the cloud space.

And so you can bet that HP, IBM, Red Hat, AT&T and a host of others have their own reasons for not being too fast in accepting AWS as the defacto public cloud. They don’t want AWS to win it all, especially when considering how cutthroat Amazon can be about gaining total control. Their APIs are closed and they can change at any time or even curtail service for whatever reason they choose.

So while Rackspace has definitely played its own games, so has everyone else.

Bias does raise questions about the promise of OpenStack and who it actually serves. I asked RedMonk analyst Donnie Berkholz about the AWS API and the question about compatibility. It’s the promise that matters, he said. API providers have a promise they must uphold. The API has to work, and that’s not a question with Amazon. OpenStack is the bigger riddle. There are so many different flavors of OpenStack that it can make it complex to integrate. Dreamhost uses Ceph for storage while Rackpsace has Swift, and Dell uses its own as well.

It’s the complexity that OpenStack has to fix. And that’s not going to be easy with so many vendors, each with their own self-interests.