Complete guide to server-based storage in its modern forms
A comprehensive collection of articles, videos and more, hand-picked by our editors
If you're an IT pro looking to implement a hyper-converged technology, you have many more options today than you had a year ago.
The list of vendors keeps growing, and the types of hyper-converged technology options are expanding as well. In this video with SearchConvergedIT, George Crump, founder of Storage Switzerland, discusses the benefits and drawbacks of several forms of hyper-convergence on the market today.
One form that vendors are increasingly offering is a software-only model that users can download and implement on their existing infrastructure. Vendors such as Maxta, Pivot3 and Nimboxx sell hyper-converged software that users can package with servers for an approach that eliminates the concern of vendor lock-in.
"The question to ask is whether your existing server infrastructure is suitable for that," Crump warned, citing potential compatibility conflicts due to storage requirements.
Another hyper-converged technology option -- reference architectures -- is one way to avoid compatibility issues. Hyper-converged reference architectures, such as VMware's EVO:RAIL, provide a "recipe," according to Crump, that gives users guidelines for installation with a similar flexibility as the software-only model.
Hyper-converged vendors are just beginning to address some of the technology's drawbacks as well. While most hyper-converged products can't guarantee performance levels for specific applications, there are all-flash models available that can deliver adequate performance across the board. Crump also said cloud integration is still a hurdle hyper-converged technology needs to address in order to be used for disaster recovery or cloud bursting.
"Some vendors have begun that move, but not as many as I'd like to see," he said.
Transcript - Hyper-converged technology options include software, flash
One of the newer options we're seeing in the hyper-converged market is a software-only model. Does that mean that customers build their own appliance, and if so how easy or difficult is that?
George Crump: One of the advantages of a software-only model is that you can leverage existing assets. So you can leverage your existing server infrastructure, as an example. But the question to ask is whether your existing server infrastructure is suitable for that. One of the typical requirements of most hyper-converged solutions is that the storage be internal to the server. If you've purchased a lot of servers that didn't have a lot of internal storage expansion, that can be a problem. So that's one of the things to assess there. But the value of the software-download model is that you should be able to deploy it right into an existing infrastructure.
Another option for hyper-converged products that we're seeing today is reference architectures. What would you say the benefits of those are?
Crump: A reference architecture is basically a recipe, or a pre-defined assembly of components. And that works really well because what you'll see is, in many cases, is vendors deliver hyper-converged Ready Nodes -- it might be a VSAN Ready Node, for example. The advantage of that is you have a recipe to follow just like if you were trying to bake bread. If you follow the recipe it means you probably have a better shot of getting it done right. The same thing applies to IT. The thing that I like about it is it still gives you the flexibility of the software-only model that you can pick and choose different server vendors; you can mix those typically as you go along.
We're also beginning to see more all-flash hyper-converged products on the market today. When should a user consider that and what are the benefits and drawbacks?
Crump: The all-flash hyper-converged solves one big problem that we typically see in hyper-converged architectures, and that is, how do I guarantee performance level to a particular application? Many of these architectures don't necessarily have a quality of service that says this application gets this type of performance -- they solve the problem by just providing great performance across the board, more so than any individual application would need. So that's one of the big interests there, is it makes sure you can deliver really good performance. From a price point, it's also a pretty affordable way to get into all-flash, because I'm not buying a separate storage controller, so all the economic advantages of hyper-converged systems apply, but they apply almost for double the value because you're not getting the double controller -- internal flash and [solid-state drives] SSDs. It's far less expensive than in a storage system, so that combination makes for a really attractive offering.
Lastly, what about cloud integration – how easy or difficult is it to integrate hyper-converged products with the cloud?
Crump: I think one of the weak points right now, frankly, in hyper-converged architectures is cloud integration. There are some vendors that have begun that move, but not as many as I would like to see. The basic theory would be that the software componentry that you have implemented in your hyper-converged architecture can run as an instance in say the Amazon or the Google or the Azure clouds. The value in doing that is you can use the cloud for DR, which would be an obvious use case. Another use case would be a cloud bursting type of situation as well.