Warning: I don’t have an answer to this question.
What’s the future of the data center control plane? AzureStack, Google Anthos, AWS Outposts, and Oracle Cloud all offer a public cloud control plane residing in the customer’s data center. I’m not talking about these services. I mean good old-fashioned VMware vSphere, Microsoft Windows Server, and Linux. The platforms that make up the majority of most enterprise IT application landscapes. Let’s first state the problem.
Keeping the lights on
I’ve talked about the 80/20 rule plenty of times on my platform. 80% of the IT budget got to maintaining existing technologies. The other 20% is spent on innovation. It may not sound like 20% is a lot. That’s because it’s not. Put plainly, CIO’s must figure out how to optimize the existing operations so that more dollars go to innovation.
Enterprise IT is hard, but it’s not brain surgery. We have a roadmap to reducing costs. And no, it’s not managed services. What I’ve discovered consistently over my career? Managed services isn’t a solution for organizations looking to invest in innovation. Most managed service providers (MSP) businesses depend on a race to the bottom. How can they provide the minimum services to make the maximum profits? The economics do not favor innovation. The MSP topic is worth a standalone blog post.
The roadmap is abstraction and automation of repeatable tasks. How much does it cost in labor to deploy a Virtual Machine from a template vs. building a VM using the OS installation media? As we’ve learned from past practices, the total cost of ownership for creating templates is much lower than the TCO for manual OS provisioning. However, we’ve hit a brick wall.
Limits of Automation
Beyond creating templates for virtual machines, we run into the fundamental problem with the private data center’s automation. The public cloud is the concept of infrastructure as code realized. Automation is a core principle of the public cloud. The data center is a kludge of siloed control planes.
Take a look at the number of storage systems inside a typical data center. It’s not uncommon to find hyperconverged storage for commodity workloads alongside mid-range, Tier0 storage system. Each one with a separate API. Add to it the network and virtualization stacks, and the staff may find themselves working with a dozen or more APIs.
Companies such as VMware with vRealize suite and Redhat with Ansible have long argued that enterprises can automate the complexity and thus reduce costs. Industry-wide this hasn’t happened. It’s not that these products don’t work. I find that it’s much more about the maintenance of automation than the capabilities of the tooling. Data center operators experience a wide range of challenges with code management and configuration management. Yet another topic for a future blog post.
Managed control plane
The drivers of reducing the cost and management overhead of these traditional data center systems present an opportunity. What if there was a software layer that abstracts all of these APIs? Well, first, that’s the goal of vRealize and Ansible. But let’s play along. What if there were a software layer that virtualized all of the underlying hardware? Like, let’s say VMware vSphere or Kubernetes. You see where I’m going. We are right back to the same problem.
It’s been a couple of years since I’ve last checked-in with Platform9. They had what I thought was a perfect idea. Platform9 would manage the Kubernetes layer, thus abstracting the complexity for the IT staff. However, the IT staff still needs to maintain storage arrays, HCI platforms, core network switches, and the rest of the IT infrastructure. Kubernetes doesn’t perform firmware updates for your VMAX array. It’s this complexity that needs abstraction.
Dell, with its PowerOne platform, believes it has cracked this nut. PowerOne has all of the automation that you’d expect. You can watch the sponsored interview we did. My main problem with PowerOne? What if your data center includes non-PowerOne systems. There’s still a business case for HCI for workloads such as VDI. Buying PowerOne doesn’t provide automation for VxRail, also a Dell product.
So, what’s the solution. I warned you. I don’t have a solution to the problem. Yet, you read the entire post? OK, I’ll give you something. Automate what you can automate. There’s value. Remember, creating VM templates reduced the speed of application deployment. Adopting VDI reduced the time of physical provisioning hardware while also reducing management overhead. It would be best if you tempered your expectations. Your data center will not look like AWS or even OCI. If you can get that mix of keeping the lights on vs. innovation to 75/25, you more than moved the needle.