The first part in this series on AWS drew comparisons between AWS and Walmart in the 1980s. We emphasized differentiation as the key component to compete with the cloud boa constrictor.
Open source has been the antidote to vendor lock-in and dis-innovation. In the cloud, open source platforms are attractive for the same reasons Linux took hold – low cost entry point and application portability. But with OpenStack sweeping CloudStack and Eucalyptus aside, the question enterprises need to answer as they consider private, public and hybrid cloud options is, “Is open source lock-in better than AWS lock-in?”
The stakes are high. The public cloud market will exceed $300 billion within four years. Meanwhile, worldwide spending on hosted private cloud will grow to over $35 billion over the same period.
In a hybrid cloud world, enterprise workloads move back and forth between private and public cloud. But the lines get blurry when private clouds are running on AWS, Google or Microsoft.
Building a massive public cloud is very complex and expensive. This is why developers use Xen and KVM at the hypervisor layer and increasingly OpenStack at the cloud control layer instead of Microsoft’s Windows or VMWare’s ESXi.
While some argued that three open source cloud stack options competing for leadership was healthy, the widespread adoption of OpenStack in the enterprise is easy to understand. Enterprises don’t want to reconcile technical differences in architecture, installation, administration, security and availability between three open source stacks. There is nary an enterprise with the budget or talent pool to build their own IaaS by downloading a distribution and standing up a private cloud themselves. As a result, they will contract with a provider in consuming a stack. And that stack is OpenStack.
Open Source Vendor Lock-in Potential is a Real Issue
Amazon has succeeded in making its API a de facto standard. As a result, enterprises will find it harder to migrate to an open source platform. Eucalyptus only adds justification.
With the ecosystem around OpenStack, it is still the early favorite. There are now dozens of companies, including Canonical, Cloudscaling, IBM, HP, Mirantis, Nimbula, Piston Computing, Red Hat and SUSE that have introduced IaaS products for it.
Such competition increases the risk that strategic features will gradually become proprietary. No one wants to write a cloud application only to discover a lack of portability. This would undermine the premise for an open source stack.
However, this is already happening. OpenStack has already bifurcated from a universal standard into different flavored versions, with each having unique implementation details. Rackspace OpenStack is different from HP OpenStack, which is different from Ubuntu which is different from Red Hat. And so it goes.
The OpenStack Foundation original intended to have one core set of cloud software that would be interoperable to avoid application dependencies. This would be the challenge to AWS dominance. However, OpenStack has so many open-ended options that early implementers have compromised interoperability. In a blog post two years ago, Rackspace director of cloud compute engineering, Troy Toman wrote, “In hustling to release a full suite of open cloud products built on OpenStack, we created some implementation specifics that were out of sync with common practices in other OpenStack implementations.”
Today, an enterprise can get seamless public-hybrid cloud integration where most of their services already run. So why would they stress over which open source platform to choose if they are also going to get locked-in by an open source tools vendor?
