October 24, 2013, 2:12 PM — CHICAGO - What's the better way to consume an open source platform: A productized version, or a pure distribution of the code?
It's a debate that's heating up in the OpenStack world, and has been played out at Cloud Connect in Chicago this week.
Randy Bias, CTO of CloudScaling and Boris Renski, executive vice president of Mirantis, are both OpenStack cloud enthusiasts who sit on the board of the directors of the open source project's foundation, and they each have their own companies looking to monetize from OpenStack as well. But their products are very different.
Mirantis this week released its own "distribution" of the OpenStack code. It's basically the open source trunk code that is packaged without a lot of proprietary components on top of it. Mirantis says the advantage of customers using this approach is that the distro will adapt as OpenStack evolves. Platforms from other vendors who also distribute OpenStack code like Red Hat and Ubuntu also interoperate with the Mirantis system.
[DISRUPTION:How the cloud is blowing up the network]
On the other side of the argument is Bias, whose CloudScaling platform is an "opinionated" version of the OpenStack code, he says.
CloudScaling has decided what sort of compute, networking and - to a lesser extent - storage components will be part of the cloud software it sells to customers to deploy on their own premises. This approach allows CloudScaling to provide deep technical support on its clouds, Bias says. Using Mirantis's distribution model, a customer may not have support for their individualized version of OpenStack that has been customized for their needs.
So which is the better approach? Perhaps the answer is: It depends. Do you want a do-it-yourself cloud that you are willing to architect, deploy and manage? Or do you want a packaged product that comes out of the box ready to plug in and play with?
Productized versions give users a platform they know is optimized for use and supported by their vendors, but they can fork users away from the trunk code. Distros can provide more flexibility, but are tied to the open source community, which can be slow to roll out new features.
If Linux is any indication, perhaps the distribution model is best. Red Hat is the most successful Linux company and it has simply distributed Linux code, in the form of Red Hat Enterprise Linux (RHEL). It's made money off of providing support for customers.
Renski says that model has worked in Linux, and it will work in OpenStack.
Bias believes that OpenStack private clouds must integrate with the major public cloud platforms that will dominate the market Amazon Web Services and Google Compute Engine. So, he's made CloudScaling interoperable with APIs from those companies.
Other companies in the OpenStack community are falling behind one or another of these camps. Red Hat, just like with Linux, has an OpenStack distribution. Red Hat OpenStack distribution is named RDO. Ubuntu has a distro that uses Canonical as its host operating system; USE has one, too.
Meanwhile, on the other side of the coin are vendors like Piston Cloud Computing Co., and Nebula, which each have productized versions of the OpenStack code.
This debate is playing out in the Hadoop open source community as well. Cloudera is a trunk Apache Hadoop distribution. Hortonworks, on the other hand, puts some of its own touches on its distribution.
Which will win out? It's too soon to tell. But perhaps there could be room in the market for more than one model, says Andy Knosp, vice president of product at Eucalyptus, another open source cloud computing platform. At the end of the day, customers don't really care whether their cloud is a productized version or a distribution of the open source code. "If it doesn't solve the business's specific needs, it doesn't matter," he says.