At Kubecon Seattle, TFiR sat down with Premal Buch, CEO of Robin Systems to discuss the evolution of Kubernetes market.
TFiR: What have been your observations so far at the conference?
Premal Buch: If we look at the Kubernetes journey, it really picked up a lot of momentum in the last 12 to 18 months. If 2016 was the year of Docker, 2017 was all about Kubernetes.
As things moved forward, we started talking about stateful sets and how to handle storage and data along with Kubernetes.
I think now the industry is ready for taking Kubernetes to enterprise stateful applications. We see a lot of discussion around that, but there are not a lot of solutions.
That’s where we come in to help customers. We have our own take on Kubernetes, which we call Hyperconvergence Kubernetes. It’s all about handling big databases for ML and AI kinds of workloads, where there are no players.
TFiR: What’s the significance of Stateful workloads in the era of stateless apps?
PB: You can run stateless applications around the higher level, but at the core of it you have data that you are leveraging when you’re running analytics or machine learning. That core data can be Hadoop, NoSQL database or traditional operational database. That database is still is handled somewhat outside of Kubernetes with more set-ups.
We change that by bringing the same portability, agility, and benefits that you get with stateless and cloud-native applications. We bring the DevOps flow and benefits to the DBA and Hadoop admins, the professionals who are dealing with this complex hairy data-driven applications.
TFiR: What kinds of customers run stateful applications, why does it matter to them?
PB: Every enterprise application, whether its operational databases running HR, finance or sales data; whether its optimizing business operations or retail experience. All the stuff that goes into big data analytics all the way to the AI and machine learning. The foundation of it is data and that data, whether it is sitting in your old school Oracle or SAP database or it lives in Hadoop, HDFS or NoSQL database, and this is one area where stateless doesn’t cut it.
The stateless flows are great for leveraging this data and creating temporary programs that spin up and tear down, but ultimately at the end of it, you have to deal with that massive amount of state in a stateful manner. That’s the market for Robin Systems.
TFiR: Which industries rely on stateful applications?
PB: We work with almost every fortune 500 company out there. We have a lot of traction in the FinTech, Banking and Insurance sectors. Industries that rely on analytics pipeline using Elasticsearch, Kafka, Cassandra on top of the data that maintains persistency.
But it’s not limited to these kinds of industries. We serve industries like travel, leisure, airline reservation bookings…any industry that deals with structured and unstructured data. We help these industries by putting that database on top of Kubernetes.
All of these big companies are looking for ‘digital transformation,’ and they’re talking about DevOps flows which are great for applications that are not data-driven.
But when they start looking at their heavier data-driven applications and try to figure out how to modernize it, that’s where the problem arises as those applications were not designed with containers and microservices in mind.
At the same time, there is a lot of assumption that’s built into Kubernetes which these applications break. There are two choices – either rewrite these applications from scratch or refactor them. Or, you will have to resort to doing things in old ways with bare metal silos.
That’s where Robin System comes to their rescue. We take all these applications run them on Kubernetes without any modification. We don’t modify the app or Kubernetes as we are 100% CNCF compliant. We just put a magic layer on top of the application, which we call Hyperconverged Kubernetes, which allows you to run your apps and take benefit of DevOps workflow.
Customers can embrace multi-cloud strategy as they can now run them on-prem and on different clouds. They can pack more applications on the same amount of compute and storage resources while ensuring the predictability that you need for the enterprise side.
It’s a marriage of both worlds – it brings the benefits traditionally associated with Kubernetes, along with the enterprise requirements like role-based access control, IOPS control and ability to use existing applications that were designed before the arrival of containers and use them as is.
TFiR: Kubernetes is a relatively new technology, what challenges do you still see users facing when they plan to embrace Kubernetes?
PB: The way Kubernetes is designed, we see two problems. One is that it’s fairly complex so a lay user who is not a Kubernetes expert often struggles to get started with it. So there is a usability aspect where people do want the benefits of Kubernetes without having to deal with all the complex scripting work behind the scenes.
The second issue most users face from a design perspective. Kubernetes embraces microservices philosophy, which allows a user to make container anywhere and all the information that you need goes with it. It’s captured inside the pod, but that design breaks with a lot of custom applications that were designed without containers in mind.
These are the problems that we try to solve for our customers.
TFiR: Where do you see Kubernetes heading in 2019?
PB: I think we have reached the point where some of the easy stuff has been done. We are now in the phase of taking care of the hard stuff – all of these enterprise applications. So far we have early adopters who like to experiment with code and run it, now we have to move towards use-cases which are more conservative. We have to go mainstream.