Hangover

In the middle of the past decade, it seemed that the leaders of the OpenStack Foundation could hardly believe the success of their product. For a long time, OpenStack trade fairs and, above all, the OpenStack Summit were surrounded by something like a mystical aura: 8,000 participants and more regularly joined the private cloud computing environment camp to catch up on the latest technology. In 2012, an OpenStack sensation began that the industry had never experienced before, especially in the open source software context.

A big bang is inevitably followed by a big hangover at some point, and OpenStack was no exception. Some participants even left the OpenStack party while it was still in full swing. Large OpenStack projects sprang up like mushrooms, fizzling out after months or years, and leaving little behind except frustrated people. Others turned to Kubernetes, which was about to become the next big thing. Following that, the media went quiet about OpenStack.

Some believe that OpenStack has disappeared from the scene, but – to paraphrase Mark Twain – the news of its demise is exaggerated. OpenStack is still an active project today, albeit with a much smaller community. However, it is still evolving, which is reason enough to take a fresh look at OpenStack and ask: What has changed technically, organizationally, and administratively?

OpenInfra

The most noticeable innovation, and one that has packed the biggest punch, is the reorientation of the OpenStack Foundation. Originally, the Foundation had formed to give OpenStack a non-commercial home, which makes sense in the US in particular for various reasons (e.g., a virtual-only project cannot hold any rights to trademarks or brands). On top of that, someone had to pay for the OpenStack party, and one of the Foundation's core tasks is to raise sponsorship money. This task is done on a corporate and individual membership basis, and the money is used to fund the (fairly expensive) OpenStack Summits (Figure 1), among other things.

Figure 1: OpenStack Summit 2022 in Berlin, live again for the first time since the emergence of COVID-19, is still a reunion of sorts, albeit a much smaller one. © OpenInfra Foundation

At the height of the OpenStack hype cycle, the OpenStack Foundation managed to gain significant influence with major vendors like VMware, Intel, Red Hat, and others. As OpenStack's relevance waned, that influence threatened to shrink again, much to the displeasure of the folks at the Foundation. A few years ago, a course of reorientation was determined: The OpenStack Foundation became the OpenInfra Foundation, and the OpenStack Summit became the OpenInfra Summit. Ever since, OpenStack has been a topic of interest, but it is no longer the only subject of interest to the Foundation.

In recent months, for example, the Foundation has successively revised many tools from OpenStack development, systematically placed them under some kind of project management, and published them. The topics tackled are something like the evergreens of the infrastructure and open source theme: Zuul, for example, the development suite that the Foundation uses to develop OpenStack, is now frequently used outside the project. As with OpenStack itself, the Foundation is responsible for ensuring that the project does not disappear without a trace.

Additionally, the OpenInfra Foundation has put out feelers in quite a few other directions. Similar to the Linux Foundation, it offers an umbrella for open source projects, although the focus is on software that is somehow used in the infrastructure sector. The log manager, Loki (Figure 2), a lightweight alternative to the classic Elasticsearch-Logstash-Kibana (ELK) stack, has now also slipped under the umbrella of the OpenInfra Foundation. Critics welcome this development because the apparent omnipotence of the Linux Foundation had gradually become a problem for many. However, it remains to be seen whether the OpenInfra Foundation will succeed in becoming a long-term and powerful counterweight.

Figure 2: Loki is a lightweight alternative to the ELK stack and is now one of the OpenInfra Foundation projects. © Loki

Technical Progress

The truism "you're always smarter in retrospect" fully applies to OpenStack. In terms of technology, many of the teething troubles of the early years are just now disappearing from the environment. Several factors play a role: Something recognizably good for the project is the waning interest of the major vendors.

Red Hat and Canonical are officially committed to OpenStack to this day, but SUSE, an OpenStack pioneer, has largely abandoned OpenStack development and pulled the plug on its own OpenStack distribution, SUSE Cloud. Accordingly, there is little coming out of those quarters in terms of OpenStack source code at the moment. However, it is apparent that the large number of companies that wanted to influence OpenStack, especially in the initial phase of the project, was probably more of a curse than a blessing. A good example of this is provided by Cinder, the OpenStack component that manages storage. When industry leaders argue about architecture decisions on the Cinder mailing list, the only way out is always the lowest common denominator. Now that fewer major players are involved, developers have occasionally dared to make major changes and features that break with old compatibility defaults.

On top of that, the OpenInfra Foundation has now found a stable work approach to help manage the individual OpenStack components and their development in a meaningful way. To date, each component has an engineering owner elected by the community from among its members. The owner is responsible for the development content and is newly elected for each OpenStack development cycle.

Few Massive Changes

If you look at the present OpenStack changelog, you quickly notice that the individual components are now being developed more thoughtfully. In pioneering days, the industry had high expectations that each new OpenStack version had to light a bonfire full of features with groundbreaking innovations. Currently, OpenStack is developing at a far slower pace. Although OpenStack releases do still include thousands of commits and a very long changelog, today the release highlights (i.e., the most important and profound changes in an OpenStack release) tend to fit on a few screen pages, mostly because new OpenStack versions no longer come with new components that no one has heard of before and then disappear into a black hole again two releases later.

The last major OpenStack release ("Yoga," numbered 25) came out at the end of March. A look at its release highlights underscores the fact that OpenStack is now more cautious than it was a few years ago. OpenStack Blazar, for example, is a relatively new OpenStack component that lets users reserve resources for themselves without using them. This ability improves reliability for applications that need guarantees in terms of resources available and the extent of these resources at a certain point in time. In OpenStack Yoga, Blazar can now store more details about individual virtual instances so that the OpenStack Nova scheduler chooses the hosts in a more targeted way.