Kubernetes and Containerization in OpenStack

Let’s take a close look at the reasons why organizations choose OpenStack as the environment to run their Kubernetes workloads:

• Cost Efficient: OpenStack users experience significant savings because it’s open source, in fact, OpenMetal customers who move Kubernetes workloads over from public clouds typically reduce their cloud bill for that workload by approximately 50%.
• Scalability: OpenStack supplies the necessary infrastructure resources to support Kubernetes scalable container orchestration platform.
• Flexibility: Unlike proprietary cloud models that may be strictly public or private, with OpenStack organizations can choose their deployment model (public, private or hybrid).
• Security: With the OpenStack code openly available organizations can find code vulnerabilities before security breaks or make adjustments to their OpenStack clouds to increase the security feature. OpenStack also has RBAC with the OpenStack Keystone project and network isolation. 
• Open Source: Both OpenStack and Kubernetes are open source, meaning that organizations benefit from the contributions of the open source community that works tirelessly to improve and advance both of these projects. As mentioned before, 85% of OpenStack deployments are also using Kubernetes, so organizations can trust that the community is consistently working to optimize the integration of these two open source projects. 

What is Magnum?

Magnum is the OpenStack project that is used to deploy container orchestration engines such as Kubernetes, Docker Swarm and Apache Mesos on OpenStack clouds. Magnum integrates with other OpenStack projects such as Nova, Neutron, Cinder and Keystone to leverage the existing infrastructure for your containers, additionally it can use templates from another project called Heat to manage the lifecycle of container clusters. Magnum uses OpenStack security features to allow users to deploy secure and isolated container environments. Organizations can use Magnum as a unified method of working with various containerized applications on OpenStack.

Why Deeper Integration Matters

With so many OpenStack users running Kubernetes workloads on their OpenStack deployments, it is clear that deep integration between these open source projects is essential. There are many tools that organizations can use to deploy Kubernetes on OpenStack such as Kubespray, Magnum and Rancher, but to make it even easier for users, some OpenStack providers like OpenMetal have turn key Kubernetes infrastructure built into their clouds allowing users to spin up clusters as easily as a VM.

Challenges and Considerations

While there are numerous benefits to running Kubernetes on OpenStack, an organization that chooses to take on this task will face some challenges. Mainly their complexity. Integrating both of these open source projects and maintaining them over time will require expertise in both OpenStack and Kubernetes. This may be too much for organizations with limited engineering talent on hand. But this obstacle can be overcome by using a hosted OpenStack provider, so you don’t need to maintain the underlying infrastructure. And it can be further simplified if your provider supplies you with a cloud that already has Kubernetes integrated. But you may miss out on opportunities to optimize your deployment if you don’t eventually bring this skill set on board. Firstly, let’s be clear, this is not a disadvantage when compared to running Kubernetes on a public cloud where you do not have the flexibility to fine tune your cloud. But one of the advantages of using an open source private cloud is that users are able to make root level changes to optimize your cloud for your unique workloads, but to do this you will need the technical expertise on hand. To take advantage of these perks, you can either invest in OpenStack training, use freely available OpenStack training resources or subscribe to a managed service from a consulting firm or your OpenStack provider. 

Kubernetes on OpenStack: Real World Example

As an OpenStack provider, OpenMetal has many customers running Kubernetes workloads on OpenStack clouds. But we’ll look at one example, Pypestream. Pypestream provides a next-generation conversational AI solution. Their customers use their chat box to provide their customers with a better chat experience, and considering some of their customers are HBOMax, Gillette, Farmers Insurance, etc. you can foresee that they have high demand. Pypestream moved workloads over from AWS where they faced high costs, unpredictable billing and competitive conflicts with customers (customers who competed with Amazon did not want their money flowing back to Amazon Web Service that often subsidized Amazon).
Pypestream moved Kubernetes workloads from their AWS public cloud instances to a private OpenStack cloud core which has 3 standard servers running their control plane. They later added several V1 servers and Ceph storage clusters (in addition to the built-in Ceph storage that comes with OpenMetal cloud cores). They are currently using OpenStack API integrations and cloud capacity to run Kubernetes across 140 compute instances. 

So what was Pypestream’s experience moving their workloads from public clouds to private on-demand OpenStack clouds?

• 50% cost savings over AWS was realized in the first 6 months, funds they were able to reinvest into their business to acquire additional personnel, resources and initiatives. These savings also allowed them to implement competitive pricing to give them an edge over their competitors. 
• Ease of Migration. This is a given since Kubernetes is designed to allow for easy migration across various infrastructure, but OpenMetal gave Pypestream the ability to use OpenStack “as-a-service” for the quickest transition path. 
• Pypestream now has greater agility within their testing and production environments which increases productivity and gets their products to market faster.