peshkova - Fotolia
Rackspace co-founded OpenStack with NASA in 2010, but its role as a contributor to the community has diminished as companies such as HP and Red Hat take more of a lead in the open source project.
With the semi-annual OpenStack Summit days away, Darrin Hanson, vice president and general manager of Rackspace OpenStack Private Cloud, sat down with SearchCloudComputing.com to discuss the challenges of custom-built OpenStack clouds, use among enterprises and the potential for Rackspace to reassert itself in the community.
You say a lot of your business now comes from do-it-yourself installs that haven't worked out. What's driving that?
Darrin Hanson: A lot of companies build something purpose-built and they get a lot support at the front end … but the way [some] companies define support is very different. It's "Here you go, here's your custom-built OpenStack cloud, call us if you need us."
What we try to do differently is build something we know we can help them operate; we know we can deliver the four nines of uptime and we know it can scale effectively because of the way we designed it.
Do you still need OpenStack knowhow even with added support?
Hanson: You have to be either developing or re-architecting applications to take advantage of the OpenStack APIs -- full acknowledgement that you minimally need to know enough about OpenStack as a consumer of it that you know how applications need to be written and how legacy applications may need to be re-architected.
How do you explain to potential customers what OpenStack is and what it isn't?
Hanson: OpenStack is in and of itself a collection of open source projects. It is not a product. It is not something that you can just deploy and have a cloud. Like a room full of Legos, there are so many configuration choices and so many architecture choices you can make. Nova alone has 700 configurable attributes, so when you take Nova and all the other DefCore projects, there are thousands of iterations of ways you can deploy a cloud.
What any provider needs to be able to come in and do, and what we do, is take the projects that are DefCore and anything we deploy in our cloud has to be deployable via our automation. It has to be able to be deployed in a way that delivers high availability and we have to have trained our people on how to manage it, upgrade it and optimize it for your environment. Once we do those things, we can deliver the plans and expertise to build a Millennium Falcon out of a room full of Legos.
What is the state of OpenStack? What has improved and what still needs to be worked on?
Hanson: The next set of projects that come into play -- Trove, Ceilometer -- these are OpenStack projects that don't yet meet our criteria for providing a [service-level agreement] around, so for us it is slowly incorporating other features and other projects that are coming down the pike from an OpenStack community standpoint.
As far as how it's getting better or is it ready for primetime … we have Fortune 100 retailers, top 10 financial institutions, gaming companies -- plenty of customers that are using this platform for production, revenue generating, customer-facing workloads. If you're developing applications correctly and if you are very clear on the strengths of OpenStack and how to develop applications for OpenStack, we are past the point of is it enterprise ready.
There are those who feel the technology hasn't matured quickly enough. What's your reaction when you hear that?
Hanson: Enterprise customers and midmarket customers are worried that it goes too fast. They're used to a three year software upgrade lifecycle and I'm more typically trying to calm my customers down about the fact that there's a six month major release cycle and a two month minor release cycle. That stresses people out, so we're spending more time talking to them about [how] we make the upgrades very seamless and very non-disruptive.
Are some of the larger vendors that have become more involved in the community exerting too much influence?
Hanson: The community is pretty diligent about rejecting code commitments that include some way for proprietary tools to have a leg up or be integrated with the trunk OpenStack.
We were once the bad guy -- the company that everybody thought was exerting too much influence --and we stepped way back, probably over to the other guardrail relative to not wanting to exert as much influence as one of the founding members. Now you're hearing that said about other companies. There are maneuvers I think Rackspace can take to reclaim a little stronger leadership position. One that is specifically focused on [and] what we have always been focused on -- trunk OpenStack -- [has] fully open source interoperability with other platforms and no proprietary influence whatsoever.
We often hear about unnamed enterprises using OpenStack that are reticent -- whether it's for competitive reasons or because they're still in development phase -- to publicly discuss their use of the technology. What's behind that?
Hanson: I don't know what that's about. It's frustrating, but what we try to tell them is you want this community to win, you want this platform to win, because that's when you'll start to receive the injection of talent and investment and things will move faster. The features you want will move faster because the community just blossoms the more there are stories of adoption.
They are starting to see that there is a win to be had both with their professional careers and for their companies to come out and start to publicize this stuff, so I hope that between Vancouver and Tokyo we have a lot more coming out of the woodwork [who] will be able to talk about it.
Trevor Jones is the news writer for SearchCloudComputing. You can reach him at firstname.lastname@example.org.