How to Embrace the Next Generation of Hybrid Data Centre Technology

“Everything started with the mainframe,” says Angell. “When the mainframe came to market, that gave birth to virtualization and containerization. And of course the micro PC market came, where we had one visible server with one application and one operating system. And then virtualization came along. Virtualization is really where the paradigm shift started. We were able to do much more with much less.”

“As technologies like converged infrastructure came around, we started to really see a division between different groups that were traditionally managed storage and applications. So then hyperconverged infrastructure (HCI) comes along, and HCI was really a big paradigm shift for how we manage it. We’ve now got this single pane of glass for managing the entire infrastructure, and it made life a lot easier. However, cloud computing came along and really changed that.”

“With the advent of cloud computing, people starting moving workloads from their traditional data centre into the cloud. Problem is, virtual servers aren’t as portable as they should be. On premises, that is not a problem; you can move workloads between different clusters, different servers with very limited time and resources. But you can’t move virtual machines (VMs) between different clouds.” And that’s how hybrid cloud became a thing.

“HCI has been traditionally just running virtual servers,” says Angell. “But a few vendors are doing really cool things to change that. Things like containerization running native applications are now becoming more prevalent. And this is what I classify as HCI 2.0. We’ve gone from traditional virtual workloads to running more agnostic cloud-type workloads.”

“Most businesses are still operating from a traditional standpoint, running virtual servers, maybe a couple of applications in the cloud, but that’s really it. Now, we really need to start thinking about the next generation of cloud-agnostic. So as these new HCI platforms come, we need to be able to run applications, whether it’s containerization or serverless functions, outside of the traditional cloud and get them on prem,” says Angell.

“Now, the benefit that containerization gives us is portability. These applications are extremely tiny and they can move freely between different environments, unlike virtual servers, which are terabytes in size and are not really easy to move between different cloud providers, or even different data centres.”

“For applications that require ultralow latency, the edge makes perfect sense,” says Angell. “If you’re in an autonomous car and you’re driving down the highway, and it decides you need to brake, do you really want that decision to be made in the cloud, which is thousands of kilometres away, or do you want it to be made locally within the vehicle?”

“AI is also going to play a role here. AI can run at the edge, but it’s going to run a lot better in the cloud. Inference at the edge is absolutely key, you can make those decision trees locally. As you are getting hundreds of thousands of transactions happening at the edge, they’re going to be making decisions on them locally, but ultimately, you’re going to want to process that information.”

“So let’s say you’re generating 3 TB of data a day from raw events. You don’t want to transmit 3 TB of data a day into a data lake, because that’s going to be time-consuming and very costly. So what we need to do is filter out those events at the edge, so you’re only sending good, valuable data up to the cloud. That can feed into your data lake, and ultimately be used by AI in the cloud.”

In his presentation, Chris Angell highlighted five new challenges that come with a hybrid data centre.

Reliability. This is no longer about hardware and software, according to Angell. “We’re talking about reliability of the applications, which in a traditional data centre, is monolithic applications living in VMs. When we’re adopting the next generation of data centre that enables Industry 4.0, we’re going to be looking at containerization, microservices functions and the whole service suite of tools that are out there.”

Security. In many organizations, there is a divide between IT and OT when it comes to IoT. “OT folks know their craft very well,” says Angell. “They understand how the production workflow operates, which is something IT doesn’t do. So we need to set guardrails around what operational technologists do, and what traditional IT folks do. Building a fabric around this, where you can have delegation of control for these various groups is going to become increasingly more important.”

Consistency. “I’m a big fan of something called one-way same-way,” says Angell. “And this leverages technologies such as Terraform for infrastructure management, so you can create playbooks that allow you to deploy VMs or applications at scale. You’re using a scripted approach. By implementing that one-way same-way methodology, you’re getting consistency across the entire board.”

Scalability. If you’re leveraging serverless technologies, you no longer need to worry about resource use, because they require next-to-nothing to boot, says Angell. “Small applications only require a few MB of RAM, and a couple cycles of the processor, unlike traditional virtual machines, which are going to run full-tilt whether they need the resources or not.”

Accessibility. This involves giving different groups within your company access to different technologies. “You want your operational technologists to focus on production-level stuff, and you want IT to focus on IT,” says Angell. “You also need to give access to security folks to get that observability as well.”