Platform9 is an awesome OpenStack-as-a-Service solution that ties into your enterprise infrastructure running vSphere. I’ve seen them present at Tech Field Day 10 and I encourage you to watch the videos posted here and read my overview and architectural deep-dive posts.

Let’s look at OpenStack specifically, as I have my doubts about the project:

I am afraid though that the OpenStack-specific use case will disappear over time if the OpenStack project isn’t able to regain focus and find the right scope. It’s not by chance container cluster management software like Kubernetes, Mesos, Docker Swarm and others have gained so much popularity recently: containers (and unikernels) is where it’s at these days.

OpenStack is fundamentally challenged

OpenStack, in the early days, was a well positioned cloud management platform. They competed against the cloud management systems offered by both public cloud vendors (who built or customized their own stack), 3rd party vendors that offer a CMP (like Red Hat CloudForms, vCloud Director and vRealize Automation, OnApp, AirVM, etc.).

OpenStack has partly failed in this battle, I think, because the product is too complex to install and has focused on integrating with legacy products too much. It generally lacks richness in features. This really means OpenStack moved from being a DevOps-minded cloud management platform to nothing more than a VMware vSphere alternative. Neither are really app- or developement-centric, but rather VM-centric and thus meant for the infrastructure administrator.

With the two architectures, VM-centric and App-centric, mixed together, OpenStack becomes increasingly complex and bland. And lastly, OpenStack has missed the Container and Unikernel boats. For a primer on Unikernels, I recommend reading up on the subject on the Xen wiki, look at this FOSDEM presentation but more importantly, have a go at it yourself with this Vagrant setup.

How does this fit in with the Container, Unikernel trends?

OpenStack is, and will be for some time, a great platform for bigger, DevOps-minded enterprises that build their own software and are able to leverage all OpenStack has to offer because their software development methodologies align with how OpenStack operates. These companies are using OpenStack mostly for their APIs, and are likely early adopters of new software development methodologies and application architecture, and thus prone to move to container and/or unikernel technologies, possibly in the public cloud, too.

I am weary OpenStack has a long future cut out for itself with smaller IT shops or enterprises that run legacy client/server (and/or Windows-based) applications. With Platform9 aiming for this install base, I’m not sure they should solely count on OpenStack becoming the standard platform for managing and surfacing infrastructure to developers or IaaS users. These organisations might move to better integrated HCI solutions, doing away with the need for Platform9 (if they’re keen on keeping the VM-centric model) or moving on to public cloud services (if they’re keen on moving to the App-centric model).

Using OpenStack for containers would only make sense for those that have OpenStack running already and want to move from VM-centric to modern App-centric development. Any organization wanting to enter the container and unikernel space will more likely look at Kubernetes, Mesos, Mesosphere or one of the public cloud alternatives.

Back to Platform9

With all this being said, is Platform9 a valid choice for enterprise to run OpenStack? Yes, I think so, for the following reasons:

  1. Platform9 makes OpenStack easy, lowering the entry barriers. It allows current IT infrastructure to be given a new life by adding OpenStack features to existing environments. This makes sense, because many enterprise IT shops with on-prem infrastructure are looking to implement OpenStack.
  2. OpenStack as an more generic IaaS Cloud Management Portal certainly has its strong suits and is very much suitable as a private cloud platform for organizations of any size.
  3. OpenStack has a firm grip on a large group of developers leaning heavily on OpenStack APIs.
  4. OpenStack is undertaking efforts to align with container and unikernel technologies, opening up the roadmap for enterprises to move from Pet VMs to Cattle VMs (implementing DevOps, desired state configuration), evolving to container and unikernel architectures and even moving on to the ‘flock of birds‘ in the longer run using the same basic platform. Mostly interesting to organizations already knee-deep into OpenStack.

Can Platform9 enable on-prem infra admins to shift to Container and Unikernel workloads?

In my Platform9 overview and architectural deep-dive posts, I’ve taken a closer look at Platform9, its use case and how their technology works to enable this use case. I started this post with a couple of fundamental issues OpenStack is facing in moving towards Containers and Unikernels.

Bich Le, co-founder and chief architect, mentioned they’ve had a lot of success with OpenStack, but they realized early-on that trends come and go, software evolves rapidly and companies that don’t adapt die.

Demo of Managed Kubernetes by Platform9

At Tech Field Day 10, they gave us a sneak peek at some stuff Platform9 is working on in the labs (check it out here: From Platform9 Labs: Toward a Polyglot Datacenter). And wouldn’t you know it, they have been working on supporting Containers (and I’m assuming Unikernel support is following suit). They’re not taking the OpenStack route here, though. In the demo, they were using native tooling to demonstrate their ability to support containers.

Screen Shot 2016-02-22 at 22.42.28This first screenshot is showing the Clarity UI, Platform9’s web portal to OpenStack. However, instead of hypervisor in a cluster, Bich was showing ‘containervisors’, or hosts capable of running containers, managed by either Docker Swarm or Kubernetes clusters.

Screen Shot 2016-02-22 at 22.43.11 Screen Shot 2016-02-22 at 22.43.04 And the CLI-view with the custom-built utility called qb.

The workflows of adding and managing clusters and hosts was, due to the flexibility of the custom-built Clarity UI, identical to the hypervisor-based platforms.

So, what Bich showed us in his demo was Platform9’s ability to support both OpenStack and Kubernetes from the same basic service. The Clarity UI is capable of addressing both these open source platforms with a single, uniform UI portal, user experience and service quality (as defined in the SLA).

Conclusion

With the architecture and delivery model Platform9 has mastered, they’re very well positioned to take advantage of each wave of innovation in infrastructure management and software development methodologies.

They’re not hopelessly tied into a single technology (like OpenStack), but are flexible enough to add new software projects to their portfolio, enabling customers to innovate easily and let them enjoy the newest trends faster than they’re able to on their own.

And here lies the real benefit of using a managed service like Platform9: a well though-out architecture, early development on new technology and the ability to innovate on a larger scale ensures customers can seamlessly move from a once-popular software stack to a newly-popular software stack.