CloudPulse Strategies CloudPulse Strategies CloudPulse Strategies CloudPulse Strategies CloudPulse Strategies

    OpenStack Uncontained: Thoughts from OpenStack Summit Barcelona Day 2

    1024 817 CloudPulse Strategies

    This post is second of two posts on the Barcelona “Ocata” OpenStack Summit. My first post, which can be found here, focuses on my day 1 impressions from the keynotes.

    For the day 2 keynotes, I expected to hear more about the project direction and strategy from the Foundation, with an emphasis on resolving some of the confusion currently seen in the OpenStack approach to containers.

    In fact, none of that really materialized. The day 2 keynotes were pretty dry, with the one notable exception being the Interoperability Challenge demo, in which 16 OpenStack vendors demonstrated their products’ “interoperability.”

    Why the quotes? Well, as impressed as I was with the demo, in retrospect, what we really saw was the vendors deploying their own products and running the OpenStack RefStack compatibility test suite. This is a bit different from what user’s likely want when they think of OpenStack interop, namely workload portability, multicloud management, etc. This would have been nice to see. It’s this sentiment that prompted Ben Kepes to ask:

    Otherwise, today’s keynotes and the subsequent media/analysts session didn’t offer much new. (Though I got a lot from the NDA customer panel at the media/analyst session.)

    In particular, my expectations for the container discussion were not met. As I’ve previously suggested, I’ve wanted the project to take a firmer stand on what is the vision and direction with regards to containers, and this wasn’t done at all.

    I’ve harped on this for a while, mostly because, prior to the Austin Summit, the project advanced a very well defined view of the role of containers in an OpenStack world — OpenStack will manage the infrastructure and containers will sit atop the infrastructure inside or beside (using Ironic/bare metal) VMs. Then in Austin, the Foundation gave the keynote stage to a group of vendors that demonstrated the inverse–OpenStack running on top of container. These vendors advanced the narrative that couldn’t be more different, i.e. that “OpenStack is just another app running on a container platform.” The stark departure from the previous message was never addressed, and I thought (and think) that the Foundation should have acknowledged this shift, the user confusion that it introduced, and advanced a cleanly articulated forward-looking strategy.

    But this didn’t happen in Austin and didn’t happen here in Barcelona. Instead, what we’ve seen here is more of the same: a collection of various demos, subprojects and use cases illustrating various approaches to containers and OpenStack that have little in common with one another but containers and OpenStack.

    In particular, we’ve seen numerous flavors of both:

    • Containerized OpenStack: The official “Kolla” subproject, the aforementioned “Stackanetes” project that was demoed in Austin, and BBVA’s presentation on how they deployed OpenStack on Kubernetes using Rancher.
    • OpenStack Running Containers: The official Magnum project, Crowdstar’s keynote presentation describing their use of Rancher on OpenStack (vs the opposite in the case of BBVA) to deploy container orchestration engines (schedulers) and containerized apps, Red Hat OpenShift on OpenStack, etc., etc.

    Until this afternoon, I saw this as a big problem, because any “normal enterprise” looking at this state of affairs is just going to check out and wait it out. But then what’s the point of adopting cloud if it creates an instant legacy environment that can’t be used with the latest agility creating and cost savings tools, i.e. containers, CI/CD, etc.

    But what I realized is that maybe it’s just me. Perhaps the OpenStack project and the foundation have come to terms with the idea that OpenStack is not for normal enterprises that want a more-or-less turnkey cloud management software, but rather they’ve accepted that it’s the domain of huge telcos and media companies deploying large scale, highly customized environments. In other words:

    In this world, OpenStack is really a product for the developers of OpenStack i.e. those large telcos and media companies, who are big contributors to the project and also are the ones promoting these various add-on projects for implementing containers. Is that what OpenStack wants to be?

    Anyhow, enough of the late night rant. All in all I enjoyed the event and catching up with my OpenStack peeps, appreciate the opportunity to attend, and I’m looking forward to seeing how all this shakes out over the next six months.

    Let me know what you think on Twitter.

    [Disclosure: Rancher Labs and Red Hat are clients.]

    More of my tweets from the event:

    Leave a Reply

    Your email address will not be published.