The OpenStack Summit in Barcelona had a mixture of nostalgia and a view into the future.
Some attendees felt wistful of the days when there were many startups creating OpenStack software, such as Cloudscaling, Nebula, or Piston Cloud Computing. Those days were like Cambrian explosion, 541 million years ago, when many life forms appeared. There was plenty of enthusiasm, large, lively parties at OpenStack Summits, and a feeling that anything was possible. Instead, the firms listed earlier have been acquired or shuttered.
That is not to say that OpenStack life forms have started to decline. Instead, large, well established systems firms such as Dell EMC, HPE, IBM, and Oracle have continued to provide OpenStack as part of their offerings, particularly as part of their hardware or SaaS offerings. GNU/Linux software firms such as Canonical, Red Hat, or SUSE have enhanced their OpenStack offerings based on their operating systems since it is a natural extension. VMware's Integrated OpenStack is in a simiar position that extends its vSphere foundation to provide OpenStack.
Smaller entrants such as Mirantis, Platform9, and ZeroStack have OpenStack-based offerings. New hyperconverged systems vendors such as Nutanix have OpenStack solutions. NFV and telco use cases have definitely moved to the forefront as telecom operators see clear a reduction in costs from NFV solutions and a desire to create a cloud to run a flexible telco infrastructure. Wind River offers a telco-grade OpenStack Solution as do many firms such as Red Hat. Telco vendors such as Ericsson (Cloud System), Nokia (CloudBand), and Huawei (FusionSphere) have adopted OpenStack as a foundation, and an open source project OPNFV uses OpenStack as its cloud foundation as well.
So the heady days of early OpenStack may be gone, but we have settled into a groove where actual uses case drive the project’s development. Enterprise or education OpenStack adoption may be coming slower than expected (we continue to see a similar set of reference customers such as CERN or Walmart as in the past summits), but we see enthusiasm from telco operators such as AT&T take its place. The smaller firms in the OpenStack market focus on delivering with tangible use cases and benefits.
The future was also on the docket for Ocata Design Summit, an event where the next release of OpenStack is brainstormed by developers. (Ocata is a beach in Barcelona.) Attending the Neutron (networking project) design sessions, I saw that there was clear focus on keeping the project focused by working on a limited set of projects called Stadium, which places discipline on the project by encouraging projects that are aligned with the core Neutron project, which is in contrast to the early “anything goes” environment.
OpenStack as a Means to an End
Another pattern that I see is that OpenStack is not being driven as an end-goal by itself. It is a means to an end, and the emphasis is on higher level parts of the stack such as IBM’s Bluemix or Dell EMC’s VCE VxRack IaaS or PaaS solutions. OpenStack resides in these system to deliver the higher level functionality, such as CloudFoundry PaaS.
End-users also have a similar rationale. AT&T likes OpenStack since it enables them to achieve their broader SDN and cloud goals. OpenStack is a means to achieving that result.
That use-case-driven view also seems to influence the OpenStack project. The collection of OpenStack projects is reaching for higher level functions in the stack, but offering items such as container support (Magnum project), container networking (Kuryr), or database-as-a-service (Trove).
Solid foundation for IaaS only?
Not everyone is a strong believer in extending OpenStack upward. Some are strong believers in making it a solid IaaS foundation, and letting services plug into the foundation. This is in contrast to the "big tent" view of OpenStack that welcomes many projects.
Canonical’s Juju Charms fits into this category, and separates services from the foundation. This makes sense since Canonical grew up as an operating system firm and advocates creating a solid foundation. Its Ubuntu system is used in the reference implementation of OpenStack.
Mirantis has a similar philosophy where de facto layers of popular services (think of Kubernetes) can reside atop a solid foundation and they believe that effort ought to be placed into making the core foundation easier to manage using modern CI/CD methods. In this view, the competition for OpenStack is the ease in which AWS, Azure, and Google Cloud manage their own services, enabling those providers to be extremely efficient.
It is not a competition of a slew of features (the public cloud providers are pretty good at delivering new features, too). Since OpenStack is an open source community, there is the ease in which anyone can propose new ideas, but that eventually encounters friction within the community-oriented process for collaboratively accepting and developing new features, which slows down development processes.
It seems logical to try to compete against AWS by delivering new features via the "big tent" of many projects, as it is fun for developers to go after new things, but end-users may be able to consume only a fraction of it, so adopting clear focus may be the best way to deploy resources.
The focus on specific use cases such as NFV, and keeping projects focused (using Stadium as an example) serves the OpenStack community well. The unlimited horizons of its early days may be gone, but that is not to say that narrowing horizons is a bad thing.