Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Hyper-converged vs. converged choice dependent on size

Though rooted in similar concepts, the question of when to use hyper-converged vs. converged infrastructure products comes down to company size and the flexibility of your environment.

Both technologies deal with the convergence of servers, storage, compute and networking for simple deployment in the data center. Converged architectures only provide guidelines for the configuration of data center components that have been pretested to work together, and users need to purchase and assemble the suggested components themselves. Hyper-convergence, on the other hand, takes this one step further by shipping all the components -- including a hypervisor and management software -- together as an appliance.

According to Ben Woo, managing director of Neuralytix, determining whether you should use a hyper-converged vs. converged architecture comes down to the size of the organization. "If you're in a large enterprise, converged architecture gives you a little more flexibility," he said. Since the user still has to purchase disparate resources, converged infrastructure is more customizable than hyper-converged. A smaller organization may be better suited to the fixed amounts of compute and capacity that are inherent to hyper-converged vs. converged systems.

But that packaging of compute and capacity together is also what allows hyper-convergence to solve some problems convergence can't. "What you want to do is buy as few infrastructure components as possible, but then scale them as out as you need to," Woo said. The nodal structure of hyper-convergence is a workaround to those cost and complexity problems often seen in traditional data centers.

View All Videos

Transcript - Hyper-converged vs. converged choice dependent on size

To start off, I wanted to get your take on what the difference between converged and hyper-converged is.

Ben Woo: For me, a lot of this has to do with size. There are so many different companies and company sizes, and ultimately both solutions work well, but it really depends on the size of the company. If you're a large enterprise, converged infrastructure gives you a little bit more flexibility because you're now dealing with discreet components that can be upgraded or tailored to your needs. With hyper-converged infrastructure, I don't like to use the word "limited," but it's much more predictable in the components that you're getting. They're both still modular, but, essentially, the hyper-converged system measures and brings together storage, network, compute and management, all in a single footprint.

When would you say a converged infrastructure would be more beneficial than a hyper-converged infrastructure?

Woo: I think a more traditional type of application. For the big type of enterprise applications or whole SAP, massive installments, they're still important because those applications were built on a more traditional model. And that's what convergence is all about: How do you take the traditional models and improve, enhance, and scale them up? You tend to find with the Oracles and the SAPs and those large enterprise apps, they tend to have a preference towards scaling up rather than scaling out, which is where hyper-converged really comes into play.

So would you say that a difference between converged and hyper-converged is scale up versus scale out? Is that a differentiating factor?

Woo: Yeah, I think that's good. Nothing in the IT world is absolute, so [there are] always gray areas, but I think that's a great way of putting it.

What are the main data center problems that hyper-convergence solves?

Woo: Again, footprint and how dense you can get it. For data centers, particularly in the small and medium-sized, what you want to do is buy as few infrastructure components as possible, but then to scale them out as you need to. For medium-sized companies, the scaling is much more predictable. "We need X number of compute capacity. We need to double that now." And usually, there is a relatively linear -- again, relatively, we're using words that are gray in the middle -- so relatively linear with the amount of storage and networking that's needed.

Hyper-convergence has made its mark in smaller environments, like remote office, but what would you say it'll take for hyper-convergence to really make its way into larger enterprises?

Woo: It does work, but the challenge you have here is in the large enterprises. The scale-out applications are typically more cloud-like, and a lot of major enterprise applications aren't there yet. So there are two different applications of hyper-converged. On the one hand, for small to medium-sized companies, it's converging into a small footprint so that you don't have to manage discreet components. In a large enterprise situation, it's all about scaling out, and the reason we haven't seen too much of that yet is because we need to evolve and transform our applications to take advantage of that scale-out environment.

The best use case in those situations is [virtual desktop infrastructure] VDI. You want to have a predictable number of users per node, and we know that VDI is a scaling issue. The more people you bring on, the more nodes you need. The more nodes you need, the more power and more capacity, the more whatever. It becomes much more predictable scale out, and we use the word "predictable" a lot because there are three components to scaling. No. 1, you want predictability when you add more nodes. No. 2, you want repeatability because you want to get the same result and deployment processes each and every time. And the third element releases scalability of the solutions.

+ Show Transcript