Two weeks ago over 7,500 “stackers,” myself included, flooded Austin, Texas for the OpenStack Summit. The event has grown massively since its inception in 2010, where just 75 people attended the first OpenStack summit – which was also in Austin.
Every Summit, there seems to be an overarching theme. This summit, the rise of managed OpenStack was prevalent.
The concept, in a nutshell, is that OpenStack is great, but it is hard. The difficulty with OpenStack at this point in time has a lot less to do with immature software, and a lot more to do with a completely new learning curve for companies’ IT departments – both on the dev side as well as ops. To combat this, leading OpenStack vendors like Mirantis, Rackspace, Canonical and Bluebox (IBM,) all have offerings where they will deploy an OpenStack private cloud in your data center of choice and manage it for you. The learning curve is essentially cut in half (if not more,) where now, you only have to learn how to leverage the power of OpenStack, rather than learning how to run it too.
This model seems like a fit for the next wave of enterprises who are looking to accelerate their adoption of OpenStack, but does it take some of the important decisions out of the customer’s hands? Let’s take a look at some of the important considerations to make when looking at self-managed versus managed OpenStack:
The Fundamental Question
OpenStack is a point in maturity where many IT leaders are plotting it on their technological roadmaps. My favorite first question to ask these IT leaders is, “Why OpenStack?.” I find you can get so much information from this one, preliminary question. Many organizations are looking to get public cloud-like functionality in their own data centres for data residency or latency reasons. This might be a perfectly reasonable use case for managed OpenStack, where your dev and ops teams are just looking for highly available, API accessible infrastructure on demand. Others see it as a competitive advantage to make all of the underlying-technological decisions from the data centre all the way up to the code. Walmart Labs, as an example, who isn’t likely to host their applications on a competitor like Amazon, chose to run OpenStack so that they can customize it to their needs and make changes to it as the business requires it. Understanding an organization’s motives for adopting OpenStack is an excellent starting point when deciding on a consumption model.
The Power of Learning OpenStack
It is common knowledge in the OpenStack community that finding OpenStack talent is difficult. Competition is high, and the salaries these folks draw is even higher. This is one challenge that managed OpenStack helps overcome. As a managed OpenStack customer, you get to leverage the deep technical knowledge and expertise that these major vendors have on staff. In contrast, it is important to look at the business value of learning OpenStack in-house. Does optimizing the underlying infrastructure and deployment/management mechanisms of OpenStack add value to your business? In some cases (like Walmart) it may.
Canonical’s managed OpenStack offering, Bootstack, allows the customer to decide whether or not they want to “take back the keys” after 6 months of managed services. This might be a low risk model to pursue, if the answer to these questions are unclear.
The Effect on Open Source
This topic may be lower in priority for some, but for those like me who are true believers in the power of open source, it is certainly a consideration. One of the nice things about the OpenStack community is the diversity of the contributors. Sure, many of the big IT vendors are among the leading contributors, but if you dig in you’ll find hundreds of small, independent contributors ranging from systems integrators, users, complementarity open source projects and startups, which adds different perspectives to the project. If managed OpenStack becomes the defacto consumption model, it could have a consolidation effect on upstream contribution, perhaps limiting the potential of the project.
The Right Mix
The one buzzword that always comes up in cloud discussion is agility. If leveraged correctly, cloud infrastructure can provide a business with the agility to make decisions quickly, iterate, and make changes based on the demands of their customers. Perhaps, this needs to be taken into consideration when making the decision on self-managed versus managed OpenStack. All of the aforementioned discussion points may not need to be decided on, on day one. Managed OpenStack might lay the foundation for a successful OpenStack deployment. It may teach us the lessons required to make the business case for learning OpenStack in-house and it may show business value in contributing upstream. In the end, my point is that both consumption models, self-managed and managed OpenStack, might have a place in your technology roadmap. It is just a matter of finding the right mix.