Open source doesn’t work the way media narratives would like it to.
You can’t see an announcement of something, then measure its immediate success in the market, and draw final conclusions about an open source effort. It’s not like the iPhone and its seemingly instant market penetration.
Few understand this better than Josh McKenty.
McKenty, CEO of Piston Cloud, is more than a talented programmer, founding architect of NASA’s Nebula cloud computing program, driving force behind OpenStack itself, and the creator of his own juggling trick, the McKenty Madness. He’s also a first-rate analyst, a long-time observer of technology trends, and something of a tech historian.
“The cloud is the third great transition,” of computing, he told me. “We had mainframe to PCs, then the Internet. Now the cloud. I think it’s that big a deal. In five years everything will look like cloud.”
To understand how cloud will develop, you have to understand how the Internet developed. It’s not what you thought.
“What made the Internet happen is we had private and public networks, joined together, running the same protocols, speaking the same language. The security challenges got dealt with in private networks first, and they matured earlier. The public network was mostly a bunch of dial-up modems.”
In other words, the hard work of building the Internet went on under the surface, for decades, before most people knew what it was. Hard things had to be done that today are invisible, like BGP, the border gateway protocol, which controls modern Internet routing.
These kinds of big jobs can only be done in an atmosphere of collaboration, many people working together, focused on the mission rather than the money or the credit.
“Look at the Linux timeline. Go from the first release to the first commercial distro to the first OEM deal to the 10% of the data center to passing Microsoft in the data center.” It’s a process that took decades, not months, not just in programming, but in education. “If you wanted to put a web server on the Internet in 2000 you should never have considered Windows, but it still had the market share.”
Cloud will be the same. Hard work needs to be done first, under the surface, before it can really explode onto the marketplace. Even after 95% of the technical work is done, education will be required before it’s accepted. Just as with Linux.
“We’re not going to get hybrid cloud nirvanas until the private cloud has the security and integration enterprises need. At that point we’ll burst out. No one wants to burst without policy management. No one wants geographic location without policy guarantees. No one wants to burst out until they have identity settled. Those features will be built on the private cloud first.”
So rather than moving to San Antonio with OpenStack, McKenty moved to San Francisco to launch Piston Cloud. Its commercial focus is security and regulatory compliance, what McKenty calls cloud audit. Piston Cloud has taken the basics of this and contributed it back to the OpenStack project, while doing proprietary builds for government, health IT and financial services, where security is paramount and buyers prefer a proprietary solution.
To McKenty, the open source part of the cloud structure needs to be as wide as possible, the proprietary parts narrow and customer-focused, in order for these hard tasks to be done.
“While OpenStack is a great framework, it’s not really a product. It’s not a stand-alone binary you can just run. To make it a product you have to make a lot of decisions – which hypervisor will I run, what will be my network model, my authentication model, my best practices for logging – they all get put into your product and make it fit for a particular environment, with trade-offs.”
Until the hard questions of cloud are settled, in other words, a cloud is not a cloud is not a cloud. Every cloud will be unique.
When companies like Microsoft or VMWare say they have a single cloud operating system, they’re claiming more for private clouds than they can deliver. Because right now, by necessity, every private cloud is going to be a one-off, McKenty believes.
Until these hard questions are answered. Which takes time.
It’s sort of a Moore’s Second Law process, McKenty admits. The effort required expands exponentially as the job grows more complex. “Every time I make an estimate I double it, and then it’s double that again when I say it. It feels like there’s two years of work, it’s probably going to take four years to engineer it, and eight years to adopt it.”
But this doesn’t fit with how media narratives are written. Reporters want a good guy, a bad guy, and an immediate resolution. Maybe Microsoft is the bad guy, maybe Rackspace is the good guy, or Red Hat. None of that is true.
What is true is that the more all these companies work together, and McKenty believes OpenStack is the best basis on which to do that work, and the broader that open source work is, the more visible the code is, the more people can get at the central questions and the sooner they’ll be answered.
“The only way to get it right is with a broad community, building implementations, trying them at scale and keeping at it.”
Not the clear answers you may have been looking for. But real life and a reporter’s narrative are of necessity two different things.