Hyper-Converged Infrastructure

Technologists are always looking for ways to simplify the IT environment and network operations so they become more efficient, and hyper-converged infrastructure (HCI) may just be the "it" technology of 2017.

To be fair, HCI was also pretty hot in 2015–2016, but industry observers say momentum is building steadily. HCI is a type of infrastructure system with a software-centric architecture that tightly combines compute, storage, networking, and virtualization resources and other technologies into a hardware box from a single vendor. That's the appeal: everything is preconfigured and ready to go out of the box, industry observers say. Another game changer? HCI can be deployed on virtual machines in as little as 15 minutes.

Gartner estimates the market for hyper-converged infrastructure (HCI) systems grew 79% to reach almost $2 billion in 2016, and projected it will reach nearly $5 billion by 2019.

HCI "is a solution that is bringing lot of the benefits of the public cloud data centers into the enterprise data centers, so it's enabling enterprise data centers to leverage commodity servers for both the application and the storage,'' says Eric Sheppard, research director for enterprise storage and converged systems at Forrester Research.

While Forrester has not yet released its latest forecast for the HCI market, Sheppard says spending on these systems in the first nine months of 2016 was in excess of $1.4 billion. "Our expectation is it will surpass $2 billion when we finalize the numbers."

Early HCI systems, or "convergence 1.0," came pre-engineered with management infrastructure to make them easier to use, "so you didn't need all these experts in the IT shop," notes Jeff Kato, senior analyst and consultant at the Taneja Group. Today's HCI systems go further, offering advanced data services such as storage virtualization, data protection, and WAN optimization, he says.

Because many IT environments are now highly virtualized, they are turning to HCI because it simplifies and reduces the data center footprint, adds Sheppard. HCI systems come with lot of automation and a lot of control at the hypervisor management level, "which means you can have teams that are less siloed." 

In terms of use cases, HCI first gained a foothold in virtual desktop infrastructure (VDI), but is now being used for a lot of general-purpose, highly virtualized business applications such as customer relationship management (CRM) and enterprise resource planning (ERP) systems, as well as for Exchange and general-purpose SQL servers, both Sheppard and Kato say. It is also being used in remote branch offices and for Web services.

Datalink, an Eden Prairie, MN-based datacenter provider that has expanded into cloud services, was bullish on HCI early on and officials liked that additional nodes for more capacity and performance can be added, says Kent Christensen, Datalink practice director for datacenter and cloud strategy. Initially, HCI comprised "north of 50%" of the implementations Datalink did, Christensen says.

Today, HCI systems scale much better than they did in first releases, and there continue to be more entrants in the market. "More people are saying the architectural concept makes sense,'' he says.

Anyone building Web-scale IT systems could benefit from an HCI architecture, Christensen says. However, in terms of how much an HCI platform can scale, "the jury's still out. There's going to be some technological limit.

"There's the vision of Web scale and this is the architecture that you can go there [with], but then there's the reality of how far you can go today versus two or three years from now,'' he notes.

Sheppard agrees that right now, scaling is not an issue. HCI provides a scale-out, server-based architecture.

"You can collapse shared physical storage down to the same highly virtualized server cluster running the applications without giving up the [virtual machine] mobility benefits you get with a [storage area network],'' Sheppard says. "You can also start with a small cluster and scale over time, rather than buying three years' worth of resources up front.''

However, performance, or predictable performance, might become a problem with too many workloads, he adds. "We're seeing an increase of [quality of service] features from hyper-converged solutions, which is helping to deal with this issue, but it's still something you need to think about."

In addition, Sheppard says he is seeing an expansion of HCI features being offered, such as quality of service (QoS) and use of flash storage technology, which he believes will help drive VM density with limited negative impacts to performance or predictable performance. That, he says, will keep the momentum building for HCI.

"It comes down to a nice mix of benefits around reduced CAPEX (capital expenditures), reduced OPEX (operating expenses), and reduced operational risk," says Sheppard. "It's pretty powerful."

Esther Shein is a freelance technology and business writer based in the Boston area.