Hyperconverged infrastructure: who needs it?
With digital transformation high on the agenda, multi-cloud on the horizon and agility and efficiency in high demand, is there anyone who can afford to ignore this fast-moving trend, asks JASON WALSH
13 November 2018 | 0
Hyperconverged infrastructure (HCI) refers to combinations of server, storage and network infrastructure combined with management software to facilitate provisioning and management: fully-architected design that also offers simplicity and greater flexibility than is associated with traditional server room set-ups. Frankly, it is a terrible term—one that obfuscates more than it illuminates—but the trend is real, and it is growing.
For Dell EMC, for example, HCI is now the biggest single growth area.
“It hasn’t taken over everything yet, as it’s in that emerging space,” said Catherine Doyle, enterprise sales director of Dell EMC Ireland.
And Dell EMC is not alone. In March, Gartner announced its forecast stating that, across the globe, “integrated systems revenue” will total $12.3 billion (€10.6 billion) in 2018—an increase of 18.4% from $10.2 billion (€8.8 billion) the previous year.
Doyle says that while the technology is not so new as to be at the avant garde of enterprise IT, it is one that is moving toward maturity.
“I wouldn’t call it cutting edge, but it is a growing area,” she said.
“As customers change their operating model from build to buy, and move to a model akin to cloud— and want to do that in-house as well—it is increasing.”
While cloud computing has seen a move away from dedicated and on-premises IT systems in the name of agility and scalability, HCI has a similar effect on systems that are somewhat closer to traditional owner-operated servers. In other words, it simplifies compute, networking and storage while also promoting agility and scalability.
“For me, and if you look at technology and why people have bought it, why hyperconverged is a hot topic, is [because] what you’re really talking about is wholly virtualised environments,” said Matt Harris, vice president of hybrid IT for HPE UK and Ireland.
Ease of use
Despite virtualisation and, later cloud, simplifying much in enterprise IT, it has not eliminated all of the challenges: resources still need to be provisioned, and virtual machine administration is required to manage the infrastructure. Arguably, virtual machine sprawl exists in the way those who built data centres had to face physical sprawl.
The answer, and a driving force behind HCI, is simplicity.
“IT organisations need to manage their infrastructure, but also manage user experience,” said Harris.
“We’re in a world now where it’s ‘click a button and get it instantaneously’. Cloud has been prevalent because it’s easy to use and had good customer experience. What it doesn’t mean is that it’s any cheaper, and there are issues around compliance.
“Everyone [eventually] has to upgrade their infrastructure, and there’s no better way to do that then hyperconverged,” he said.
Andy Brewerton, director of channels for western Europe and sub-Saharan Africa at Nutanix, also makes the case for HCI as a radical simplification.
“The purest definition of hyperconverged infrastructure is the compression of the technology stack,” he said.
“The best way of visualising it is, if you look at a traditional setup, you have servers for your compute, storage and a network—a SAN or whatever. That’s a traditional three tier network [whereby] you have a design that is correct on day one because you’ve designed it for a workload with the right amount of compute, right amount of controllers and right amount of storage.
“In that model you might have multiple manufacturers and supply chains [and] they may talk to each other and they may not, but what you intrinsically have is complexity: complexity in purchasing, complexity in deploying, complexity in operations,” he said.
This is a problem, he says, because the designed perfection disappears due to the complexity introduced by growth.
“That’s the problem that hyperconvergence is trying to fix, and the way hyperconvergence fixes that is to shrink those lawyers down to one and do everything in software.”
This emphasis on simplicity goes right down to the metal.
“Our solution is fundamentally software, and it sits on commodity hardware: we don’t modify the hardware, we just run on it,” said Brewerton.
Dell EMC’s Catherine Doyle says that ease-of-use also means breaking away from the traditional plumbing problems associated with enterprise IT.
“Another big area is patching: you patch one bit of infrastructure, and [then] you break something else. It’s a never-ending cycle,” she said.
The traditional model simply does not work in today’s business world, she says.
“It needs to be quick, it needs to be agile—especially in a DevOps environment. Well, hyperconverged arrives and is ready to go: just plug it in.”
One factor that cannot be ignored is the emergence of HCI in a world dominated by digital transformation that is driving a multi-cloud strategy in many businesses. So what applications are being moved to HCI?
Rory Choudhuri, solutions marketing director for EMEA at VMware, says that, in fact, hyperconverged not only brings home similar benefits to cloud, it also works well in environments that already make use of the cloud.
“People are saying, ‘I want those cloud-related benefits of flexibility and scalability— particularly scalability—but I want do it in-house without having to re-engineer my processes’,” he said.
As with the adoption of other enterprise technologies—whether the initial move toward virtualisation or, later, the cloud—an initial rush has been followed by more sober growth.
“[Like] with cloud we saw an overall mad rush, but it has tempered. However, we’re still seeing a move toward it,” said Choudhuri.
This rate of growth is driven not by technology, but clearer assessments of business need, he says.
“With customers, mostly enterprise—financial services and that sort of thing—I have found [that] there is an expectation that a hybrid approach is the way things will go. There’s going to be a long, long tail of stuff that stays on-premise, and some stuff that never goes.”
Examples of customers using hyperconverged infrastructure from VMWare include not only financial services companies, but also the British Army, Israeli television and the Libyan postal service.
“We do see the whole of hyperconverged infrastructure across the board,” he said.
“About 63 to 64% of use cases are tier one applications: ERP, SQL, and a huge amount of VDI of course because it has a lot of storage requirements.
“A lot of hyperconverged infrastructure is going into those traditional server rooms; it’s just bigger building blocks [however] we do see a transition from CapEx to OpEx. There is absolutely a market for people building data centre the old way, but there are also cases where someone needs to install a small cloud at a brand outlet. [With HCI] all you need is someone who can plug in the power and the network cable.”
Dell EMC’s Doyle says that, as with cloud, some customers have gone all-in for hyperconverged infrastructure, but the more common approach is hybrid in nature.
The driving factor in both cases is the promise of scalability and flexibility.
“A lot of the fintech are going hyperconverged all-in. If you’re born in the cloud it makes sense, and you can flex in and out of the cloud. It makes that whole multi-cloud [infrastructure] much easier,” she said.
“The key is, ‘what is your application, what does your application need, and what’s the best way to service it?’ Sometimes that’s the public cloud, or low-end products. Sometimes it’s built it yourself and sometimes it’s hyperconverged.”
In this mix HCI is defining a role for itself, she says.
“Things like payments platforms, anything that needs very high performance and low [tolerance of] risk, in the sense of not having downtime, [for example] utilities have gone into it in a big way. And companies doing software development, which can be in any industry, especially those doing ‘extreme programming’.”
Nutanix delivers HCI itself as wells as via OEM with partners including Dell EMC, Lenovo, Fujitsu and, interestingly, on IBM’s Power architecture—so it is not just x86.
Brewerton says that this helps to support special use cases and promotes the goal of flexibility.
“Five years ago, the workloads you’d put on a HC would be VDI and general purpose server infrastructure. Now, over 50% of use cases are enterprise applications. We’ve seen a shift such that any application can run on HCI. We run oracle, we run a lot of the ‘new world ‘distributed applications, we do a lot of container work with Kubernetes. We don’t tie anyone to a hypervisor; we obviously allow them a choice of tin, but we also allow a choice of hypervisor.
“We’re big believers in hybrid or multi-cloud: we don’t think that cloud stops at public and private. We would argue your branch offices are little clouds—edge—and there’s also the internet of things, which is a really distributed cloud whether it’s an automated car or an oil rig,” he said.
Ultimately, says Harris of HPE, HCI means fine-tuning enterprise IT through a workload emphasis: the trade-offs are cost versus resiliency and performance expectation.
“Hyperconverged infrastructure is just a method of building the infrastructure. For the users it’s easier; you start with simple building blocks and more time using it to your benefit, freeing-up resources and spend more time thinking about how they want to grow strategically. Your KPIs are either, you’re saving the business money, or you’re helping the business grow through technology you’ve deployed,” he said.
“As for where it’s going, most organisations have almost virtualised the storage and compute [so] what we’ll see now is the virtualisation of the network.”
|HCI in the real world
Hyperconvergence also means bringing the power of compute outside the traditional office environment, says Matt Harris, vice president of hybrid IT for HPE.In at least one case this means bringing it to some surprising locations.
“One customer for us is [Formula One team] Aston Martin Red Bull Racing. They key for them is speed, agility, the differentiation you get by being able to compute there and then,” he said.
“What Red Bull Racing wanted was to be able to scale immediately, have great performance and have operational responsiveness—[and] that architecture moves around the globe as they go to new races.”
In practice, this means that Aston Martin Red Bull Racing builds a trackside infrastructure, effectively a mini data centre, and also communicates with datasets located back at company base.
“They can send information back and forth to make decisions right now, as well as take all of the logs from a race and analyse it back at base.Likewise, VMWare has seen HCI benefit industry by bringing computation out into the field.
“The datacentre is being encapsulated,” said solutions marketing director Rory Choudhuri.“A classic example is train locomotives or planes, but there are plenty more.
“You need that processing capability, but it’s not a replacement. [And as well as processing and using it on-site] you’re still moving the data back to the centre of the network. It does not replace, it simply augments.”