Is OpenStack best deployed as a server distribution, a service from a cloud provider, or something else?
You might think that such a question would have a simple answer. Yet at the OpenStack Summit in Paris last month, seven developers participating in a panel discussion failed to reach a consensus.
One reason for the debate over deployment methodology is the lack of any clear product designation. OpenStack is more an entity than a product.
“The reality is that OpenStack is not a product. It is a set of open source tools that rely on many other open source tools to populate function. There is no one thing that you can download and say you have OpenStack up and running,” said Jesse Proudman, founder and CTO of Blue Box, and one of the seven panel participants at the summit.
OpenStack is an open source, private cloud alternative to Amazon Web Services and other public cloud platforms. The software controls compute, storage and networking resources throughout a data center. It is managed through a dashboard, command line or via the OpenStack API.
OpenStack, which began in 2010 as a joint project of Rackspace Hosting and NASA, now is managed by the OpenStack Foundation. The Foundation claims nearly 10,000 individual members from 100 countries and 850 different organizations. It secured some US$10 million in funding for the OpenStack cloud computing platform.
That notion of OpenStack being just a thing makes more sense than it initially might seem, according to Scott Sanchez, vice president of strategy at Metacloud.
“OpenStack by itself is not a product companies are going to consume. It is a framework or kernel that allows companies like Metacloud or Red Hat or someone else to package it up into something that is a desirable product by the market,” Sanchez told LinuxInsider.
Some people refer to OpenStack in a similar way as the Linux kernel. People build around it. They turn it into an offer and bring it to market to serve a particular customer segment, he explained.
Ultimately, it depends on the company’s goals and needs. You can not think of OpenStack as a product like VMWare, for example. OpenStack becomes a product when a company or developer consumes the open source code, invests in it in some way, puts support around it, adds features to it, or gives it a delivery model and then offers it commercially. At that point, Sanchez said, it becomes a product.
That is a good thing for OpenStack users, Proudman told LinuxInsider. It means you must rely on the vendor community largely to come up with ways to consume the offering.
“Distributions were the initial methodology, because without having production-ready code, it was hard to deliver it as a service,” he said.
The different viewpoints about OpenStack being deployed better as a distribution or as software service are part and parcel of the debate over using a public or a private cloud service, suggested Nick Barcet, vice president of products at eNovance.
The customer’s use case will determine which makes more sense. For instance, banks and healthcare have their own requirements. So does telecommunications.
“If you want to use a public cloud, OpenStack as a SaaS is what you need. If your usage drives you to having something more private for yourself, then you would deploy OpenStack on your own computer,” Barcet told LinuxInsider.
New technology enterprises with their own SaaS offerings want to be in control of their own infrastructure to ensure they can meet service level agreements, he noted. What’s really at issue is how people make intelligent choices about what to use.
The delivery debates should change now that the concept of OpenStack as a commercial offering is taking root. The technology is advancing beyond four years and has production-ready code.
“The market has now accepted that OpenStack is a thing, and it works. Now we can all move on to how will buyers adopt OpenStack,” said Proudman.
One reason the summit panel failed to reach a consensus is that its participants were all vendors selling their solutions a specific way — each arguing that their methodology was correct, said Proudman. Initially, the market bias was to follow the methodology of the traditional open source software. That was to package something up as a distribution like the Linux kernel or an open source project.
“At the end of the day, my perspective is that the type of buyer will dictate what the appropriate production methodology for OpenStack is,” said Proudman.
Today’s vendor market falls into four buyer categories: software distribution, hardware appliance, consulting engagement, and software service. Each one fits a different use case, noted Proudman.
The largest enterprises at a Fortune 1000 firm have the operational expertise to take OpenStack as a software tool and run it internally. For that type of customer, a distribution might make sense.
“What we have largely seen is those buyers end up going to the consulting method. For everybody else in the marketplace, they want the benefit of using OpenStack. They want API, but they do not want to figure out all the ugly internals of how to run it, and how to fix it when it breaks,” said Proudman. “That does not bring any value differentiation to their organization.”
OpenStack delivery “as a Service” solves that headache for those customers, he argued. They get all of the benefits of a private cloud. It is a protected environment that belongs specifically to them without the operational burden.
All of the debate over OpenStack deployment methods springs from one major reality. The debate about everything going to the public cloud was started by Amazon more than 10 years ago, according to eNovance’s Barcet.
“The reality is that we are seeing hundreds, if not thousands, of customers using private clouds today. The public cloud landscape does not fit their needs,” he said.
OpenStack is not just one successful entity. It is a collection of entities that soon could overpower Amazon. In fact, following the Paris Summit, Barcet predicted the number of nodes deployed with OpenStack as the orchestration layer on top one day will equal the number of nodes that Amazon has.
His prediction is also based on numbers. Seeing the number of people starting to go from test deployment to production deployment and seeing the sheer size of the OpenStack deployments is amazing.
“A collection of small, medium and large deployments on private or public clouds will all be running the same software. OpenStack will be bigger than the current giant,” said Barcet.
Still an Options Game
MetaCloud’s Sanchez has argued on both sides of the distribution vs. public/private cloud debate. He previously was at a company that provided a downloadable distribution of OpenStack. Metacloud delivers OpenStack as a Service — with a twist.
Its model is a public cloud as a core OpenStack, but it’s delivered as a software service as well. This is sort of a hosted multitenant model, said Sanchez. Metacloud fits squarely in the middle of the options.
“I see customers realizing that what they are really looking for is the operation. The way the cloud version is run is attractive to them. There is a tremendous amount of service in using it through either a public or private cloud,” he pointed out.
A very small percentage of companies are pushing the envelope of technology. For them, getting a traditional distro of OpenStack, applying a lot of customization, and adding a lot of intellectual property to it probably makes a lot of sense, Sanchez noted.
“For most companies, they do not want their core competencies to be running OpenStack. They want to be building their business apps. For those companies, getting it as a service through either a private cloud or public cloud really does make a lot more business sense,” he said.
Forest or Trees?
The flexibility emerging around OpenStack’s development is a feature rather a design flaw, suggested Sanchez. At Rackspace, when the team first created the idea of an open stack, its participants knew that the opportunity was not just to go build public clouds with the software.
Rather, the software was to service companies that were using VMware and all kinds of other first-generation cloud or virtualization technologies. It was built with flexibility in mind.
“That involved more than just how they might package it and distribute it. It also involved how you might delivery it and consume it and everything else,” explained Sanchez.
OpenStack’s developers include vendors that are productizing OpenStack. Not all of them have the commercial product perspective, though. Instead, many of them are very much open source people.
“That is part of why there is a wide range of views from vendors on what to do. It becomes hard for them to see the forest for the trees. There is the open source development contributor. There is a product commercialization contributor. Their perspectives are going to be very different,” Sanchez observed.
Why the Options?
OpenStack’s list of deployment options does not constitute marketing overkill, said Blue Box’s Proudman.
The software’s flexibility is a sign of maturity that allows vendors to provide customers with a new way to actually consume OpenStack, he pointed out. This new way has a service approach.
“Some of the options fit specific use cases, but the majority of people who will use OpenStack will benefit more from purchasing it as a service versus buying it as a software distribution or as a consulting engagement,” Proudman reasoned.
The goal of OpenStack is to have the software installed in as many organizations as possible. The “as a Service” option enables the majority of customers. The other available options enable larger-sized, though fewer, customers, Proudman explained.
“If we can get a huge number of people using OpenStack,” he said, “then we all win.”