Question of the day: What should OpenStack do with all those eager contributors? Does that mean expanding features or focusing on a core?
In the last few months, the OpenStack technical leadership (Sean Dague, Monty Taylor) has been introducing two interconnected concepts: big tent and levels.
- Big tent means expanding the number of projects to accommodate more diversity (both in breath and depth) in the official OpenStack universe. This change accommodates the growth of the community.
- Levels is a structured approach to limiting integration dependencies between projects. Some OpenStack components are highly interdependent and foundational (Keystone, Nova, Glance, Cinder) while others are primarily consumers (Horizon, Saraha) of lower level projects.
These two concepts are connected because we must address integration challenges that make it increasingly difficult to make changes within the code base. If we substantially expand the code base with big tent then we need to make matching changes to streamline integration efforts. The levels proposal reflects a narrower scope at the base than we currently use.
By combining big tent and levels, we are simultaneously growing and shrinking: we grow the community and scope while we shrink the integration points. This balance may be essential to accommodate OpenStack’s growth.
UNIVERSALLY, the business OpenStack community who wants OpenStack to be a product. Yet, what’s included in that product is unclear.
Expanding OpenStack projects tends to turn us into a suite of loosely connected functions rather than a single integrated platform with an ecosystem. Either approach is viable, but it’s not possible to be both simultaneously.
On a cautionary note, there’s an anti-Big Tent position I heard expressed at the Paris Conference. It’s goes like this: until vendors start generating revenue from the foundation components to pay for developer salaries; expanding the scope of OpenStack is uninteresting.
Recent DefCore changes also reflect the Big Tent thinking by adding component and platform levels. This change was an important and critical compromise to match real-world use patterns by companies like Swiftstack (Object), DreamHost (Compute+Ceph), Piston (Compute+Ceph) and others; however, it creates the need to explain “which OpenStack” these companies are using.
I believe we have addressed interoperability in this change. It remains to be seen if OpenStack vendors will choose to offer the broader platform or limit to themselves to individual components. If vendors chase the components over platform then OpenStack becomes a suite of loosely connect products. It’s ultimately a customer and market decision.
It’s not too late to influence these discussions! I’m very interested in hearing from people in the community which direction they think the project should go.