I really do not see hyperconverged infrastructure (HCI) as the final product for the leading vendors. It is more of a minimum viable product than a finished solution. Converging compute and storage has made infrastructure management easier, but there is still a whole lot of IT that is not easier. I sometimes tell people that their role is changing or going away after they deploy HCI. I usually tell them to move up the stack. Up the stack is closer to the business that employs them and uses the IT. Up the stack is getting closer to the applications and the data, and then to the users. The next move for HCI vendors is to move up that stack to make application development, deployment, and operation easier.
Hyperconverged infrastructure as we currently have it has automated or eliminated many infrastructure tasks. A whole swath of storage-related management has gone away. There are no LUNs to configure, no Fibre Channel zoning, no storage tiering to manage, no balancing VMs across datastores for performance or capacity reasons. You simply determine the required capacity and whether you can justify (afford) an all-flash configuration. An HCI platform can support thousands of VMs without the need for a dedicated storage team. If you are all in with HCI, then you may not even need a backup team. Backups are built into most HCI platforms, along with replication to another data center for DR. All this simplification is around low-level infrastructure: essentially hardware abstraction, pooling, and management. The remaining horizontal move is to manage the network. We will see HCI vendors moving this way over the course of 2017. Nutanix has started by adding the Plexxi console to its Prism management tool. We will likely see partnerships between HCI vendors and network vendors. The HCI system may direct the configuration of the network hardware.
I believe that the next phase in hyperconverged infrastructure will be a real private cloud: not just a way to deploy VMs faster, but an infrastructure for new applications. I think it will resemble AWS far more than vCloud Director. Existing HCI delivers the ability to create VMs rapidly and supports existing application architectures. I think HCI is going to enable newer application architectures by offering more application services. These services will enable applications to be assembled rapidly. We have already seen Nutanix offer file services that are deployed from their console.
- The next logical step is an object store that is S3 compatible. File and object services are relatively simple things to build on top of the scale-out storage that is part of every HCI.
- Then containers, probably in the form of an automated Kubernetes deployment. For many of the enterprise uses of containers, it makes a lot of sense to put containers inside VMs. We see VMware with a couple of options for “containers as VMs,” and we will see other HCI vendors putting containers in VMs. Containers will form the basis for other scale-out services, like web server farms. I don’t expect these to be built into an HCI; I expect the HCI to integrate with external container image repositories.
- There will also be a role for more stateful services, like Databases as a Service. These services will be needed to store persistent information for container applications.
- Next, I expect a “serverless” platform that looks a lot like AWS Lambda. To be useful, this service will also need some queuing and load-balancing services.
- I would also like to see a whole lot of network functions: load balancers, routing, and DNS. These tie in with managing the physical network devices as a great HCI development direction.
Look at the most popular products that AWS offers, and I think you might have an idea of what I hope to see. Many of these things already exist as open-source projects; OpenStack is probably a great example. These open-source projects are just that: projects. They aren’t ready for enterprise customers to simply install and start using. This is where the HCI vendors have value to add, by making deployment and operation easy. One leader in this ease of use is Platform9. It offers OpenStack, Kubernetes, and Serverless as managed services. I wonder whether we will see Platform9 acquired by an HCI vendor in 2017. I would not be surprised.
The big value of hyperconverged is not in the hardware convergence; rather, it is the simplification that brings the most value. I fully expect to see the leading hyperconverged vendors turn their attention further up the enterprise IT stack.