Sramana Mitra: Let me double-click down on one more layer based on what you have said. Is Portworx sitting on top of the traditional storage players? Is it like a caching solution on top of the existing one?
Mohan Thirumale: Yes, it’s not unlike caching, but it’s doing two things. One is exactly what you describe. It’s a software layer that sits on existing storage that people have.
When I say existing storage, by the way, it could be cloud drives, EDS, Google storage, or storage in the Azure Cloud. It could also be service-based storage or SSDs and flash that is embedded inside of storage. Our customers are using all three of them. They continue to make that investment.
What Portworx does is, it sits on top of it and it does two things. One thing it does is it virtualize all of that underlying storage. We have an auto-discovery mechanism where we’ll go in and fingerprint all of that storage. We know to turn it into one large pool of storage.
The second thing we do is allow virtual volumes from that pool of storage to be carved out and attached to a container so that now each container has its own little virtual volume. It travels and it’s like your own suitcase. That data travels with that container no matter where it goes. It could go from node to node. It could go from one server to the other.
We will always ensure that the data is there. The benefit to the customer is several things. They could deploy these containers anywhere and they are going to get data access with high availability, disaster recovery, and backup.
Notice that the words that I’m using are the same things that age-old benefits of storage offer. Storage makes data highly available. It makes it reliable and resilient to disasters or failovers. It backs up data so that it’s available when you need to recall stuff for compliance reasons or other reasons.
We are making all those things available but for a containerized set of applications that they couldn’t do before with the old-style storage. You can view it as an extension of Kubernetes or you could view it as an overlay on top of storage. The reality is that it’s both; it marries those two.
Sramana Mitra: The concept is very much a caching kind of principle.
Mohan Thirumale: Yes, it’s like a caching layer except for the fact that we do a lot more than caching because it’s only one part of what we do. One of the key things that we do is, we allow control all of the data. We hand all the data reigns over from a storage admin to the new breed of DevOps admin – the people who are operating Kubernetes.
This is very significant now. Let’s say you are driving a retail application like a search application. They would put all these applications in the containers, and they deploy them. All of a sudden, they now use Kubernetes to orchestrate that.
The Kubernetes operator all of the sudden now can, if they deploy Portworx, provide HA, backup, DR, and all of that using Kubernetes. It’s almost like Kubernetes is a steering wheel for containers.
Think of the radio stations. You could get all of that on the steering wheels. We put all the data controls on the steering wheels for Kubernetes. The person who is operating us is not the storage admin who was familiar with the storage and the hardware. It’s someone who doesn’t know much about storage.
They are just using Kubernetes and they say, “Hey, I want this data available here. I want this backed up every hour.” They just tell Kubernetes what they want and we make that translation. The dramatic effect it has is, it takes this away from the old model of service tickets and service requests.
In the old world, someone who wanted to deploy an application would ask for computing. VMWare changed that because they could make it easy. They could just ask the VMWare, but storage didn’t change.
They had to ask the storage admin, “Hey, could you provision me 30 terabytes? Make it highly available and back it up every hour.” The storage admin would get back to them two hours to three days depending on how the system works.
When you are running real-time systems like Gmail, you go to some part of the world. You log in to the world. There are local Gmail instances that have access to your data. That cannot be done unless it is all automated. What we have done essentially is to allow self-service automation of storage and data to be available to an application administrator and remove the whole service request model.
That is why it is hugely popular. That is why people are deploying because the ability to get agility is driven by automation. What we have done is turn it into self-service automation of data.