CIO, Cloud, Data Center, Opinion

PanelCast 2019 Predictions: Hyperconvergence and the Next Wave of Systems Integration

Hyperconvergence has become a force unto itself. What comes after HCI? Is it extensions that turn HCI into a broader platform or is it a whole new architecture altogether?

This blog post was created following ActualTech Media‘s inaugural PanelCast event, held in December of 2018.  This event addressed 2019 enterprise IT predictions in a discussion moderated by Scott D. Lowe, along with four industry experts, including Sirish Raghuram of Platform9, Theresa Miller of Cohesity, Mike Wronski of Nutanix, and Jeff Ready of Scale Computing.

If you’d like to watch our very first PanelCast, please visit https://youtu.be/lbFfTbdztn8


Panelist responses to this audience question:

Taking into account the evolution of technologies, HCI seems to be at the forefront for now, but what is that going to look like in 2019 and beyond?

When you consider the ultimate goal of homogenizing the experience across all services and platforms, that is what HCI is likely to become. According to panelist Mike Wronski (Nutanix), It’s going to be “the true version of software-defined everything in a full stack.”

Jeff Ready (Scale Computing) believes that the reality of software-defined everything will drive hyperconvergence. “As you start to move to everything to be software defined, that opens up the opportunity to do a lot more orchestration and automation.”

While hyperconvergence is clearly the future, it’s still a ways off for some organizations, says Teresa Miller (Cohesity). “Not everyone has the budget to install it today.” Either way, it’s good that the conversation has started now, so that companies will be prepared once they are ready to take the leap.

Scott's Take

Really, we’re already seeing the beginnings of what a post-HCI world looks like.  In the beginning of HCI, we saw software-defined conglomerations of the hypervisor, compute, and storage layers spark something of a transformation in how people thought about their data center environments.  Today, we’re seeing HCI what I see as three ”blurry” paths for hyperconvergence.  I say “blurry” here because these are not mutually exclusive options.

First, I see HCI expanding beyond the confines of the central data center and expanding to consume the edge and, to a degree, the public cloud.  However, the public cloud component is typically handled through either some kind of overlay or as an extension to on-premises HCI environments.  This edge-centric vision of HCI makes a lot of sense for a number of reasons.  As we look at the common retail edge use case, it’s clear that individual stores can’t be staffed with hordes of IT support staff and they often barely have room for inventory, let alone a rack of servers.  With a hyperconverged solution, some of which can operate with as few as two nodes, each retail location can enjoy a robust hardware stack that can ensure that the site’s operations don’t get impacted due to hardware solutions built without availability in mind.

Second, early incarnations of HCI were touted as “linearly scalable”, which is a good thing.  However, the challenge that some organizations saw was that they were essentially forced to buy more hypervisor licenses each time these needed more storage.  Or, they had to buy more compute even if they just needed storage.  Today, we’re seeing interest in a segment of the HCI market that takes a disaggregated resource approach.  You might be tempted to think of these as composable systems, but they’re not quite there in all cases.  Regardless, the benefit of this disaggregation is that organizations can granularly size and scale individual resources to meet the demands of their applications.  If they need more storage, they simply add a storage block.  The HCI part comes in at the management layer where the resources are still managed in a singular way.  There is debate in the industry around whether these should be called HCI, but my personal take is that the terms we use to describe things need to evolve with the technology, so I’m comfortable putting these solutions into the HCI bucket.

I hinted at composable infrastructure in the previous paragraph.  This is the third path that HCI has taken and it’s sometimes considered the next evolution of HCI, but with some key differences.  In general, composable systems can support bare metal, virtualized, and containerized workloads, making them suitable for any application. HCI doesn’t usually support bare metal since the hypervisor is so integral to those environments.  Composable infrastructure, as a term, has gotten a slow start, but the outcomes – namely, the ability to weave together any combination of resources on the fly to support an application – are powerful.