Hyper-converged infrastructure options simplify virtual environments
A comprehensive collection of articles, videos and more, hand-picked by our editors
A hyper-converged infrastructure tightly integrates storage, compute, networking and server virtualization resources in the same box, and now products are starting to offer more points of differentiation.
Arun Taneja, founder and consulting analyst at Taneja Group in Hopkinton, Massachusetts, surveyed the hyper-converged product landscape in this podcast interview. He explained the distinction between hyper-converged and converged systems, updated the list of products that meet his definition of hyper-convergence, discussed the latest choices users will find for hypervisors and hardware, and offered his predictions on the direction hyper-converged storage products could take.
What sets apart hyper-converged storage or a hyper-converged system from a converged infrastructure?
Arun Taneja: Converged systems, or converged infrastructure, take the fundamental pieces of compute, storage, networking and server virtualization, possibly from different companies, and bring them together and make them operate as closely as [they] can as a whole. In other words, the vendor that will put that converged infrastructure unit together will perhaps put a management layer on top of it and a variety of other things to make the deployment and the management of those four pieces easier.
Hyper-converged storage or hyper-converged infrastructure or hyper-converged systems -- different terms being used meaning the same thing -- essentially means that one vendor has taken it upon [itself] to build an entire infrastructure and put it in a box. So, it not only has compute and networking and storage and server virtualization in a box, but it would also have other elements like backup software, snapshot capability, data deduplication, inline compression and WAN optimization. [One vendor] takes all the various pieces that make up a traditional infrastructure in the market today and put it in a single box. That is very different than converged infrastructure, which is really three or four separate pieces that are brought together, but [each piece is] still distinct in its own right. In a hyper-converged environment, you don't see any seams. It is entirely seamless.
What products fall into this category of hyper-converged storage?
Taneja: By my definition, there are three products available in the marketplace today that meet the criteria we've established for hyper-converged storage. You have Nutanix, which is considered to be the granddaddy of hyper-convergence because they were the first to market, and they are the pioneer. Then you have SimpliVity and Scale Computing, which is really focused at a much lower level of the organizations. I also think in terms of VSAN or Virtual SAN from VMware. Even though it may or may not meet all the criteria of hyper-converged storage, I think from a customer perspective, [it] would be viewed as hyper-converged storage.
But I can also say that almost all the major vendors that have servers, storage and networking all available as part of their total arsenal, they're all looking at producing something that would be considered hyper-converged storage. So, over time, I think a lot more possibilities will exist, but at the moment, I consider the four that I mentioned to be prime examples of hyper-converged storage.
How much choice do users have these days with hypervisors for their hyper-converged storage?
Taneja: The starting point for Nutanix, for example, was VMware, so that is the predominant installed base for Nutanix today. But Nutanix also made an announcement that they now have Hyper-V-based offerings. SimpliVity also started with VMware, and that is fundamentally their offering today. I do expect that SimpliVity, like Nutanix, will add additional hypervisors, but whether it ends up to be KVM or Hyper-V, I don't know.
Then the third one, Scale Computing, actually is distinct in that they were KVM based from the very beginning for very good reasons -- they are targeting the lower end of the market, which is very price sensitive. KVM, of course, is open source and gives them that lower price point. And Virtual SAN from VMware, by definition, is VMware only. The customer will have one choice.
Is hyper-converged storage generally associated with commodity hardware?
Taneja: The answer is yes and no, and let me explain why. Today, if I look at it, Nutanix is based on commodity hardware. Scale Computing is based on commodity hardware. And SimpliVity is based primarily on commodity hardware, but they do have a very specialized ASIC that is designed for very fast inline data deduplication. SimpliVity obviously determined that the only way to do inline data deduplication without impacting the application performance was to actually have specialized hardware to go with it.
Today, the answer is a mixed answer. However, as Intel compute capability continues to increase, and we've seen Moore's Law apply for the past three decades, why wouldn't it apply moving forward? At some point in time, it would not be a big deal to really see all hyper-converged storage be commodity based because, ultimately, I think that is the objective. One of the primary reasons a customer would want to go to hyper-converged storage is lower prices that are demonstrated by commodity hardware.
In what ways other than hardware do hyper-converged storage vendors differentiate themselves from each other?
Taneja: Look at the starting points of these different vendors, and you will see where they have done exceptionally good work. Take, for example, SimpliVity. The roots go back to data deduplication for them. They kind of started with the premise that when the data is created, the data should be deduplicated and then kept in that condensed format through its entire lifecycle. That became a hallmark of their differentiation. Granted, they did that with the specialized hardware, but they excel in that category.
Some might have better WAN optimization between two separate sites. Someone else might have better replication capabilities. Others might have better snapshot capabilities. All of these hyper-converged storage vendors produce what I call [virtual machine] VM-centric solutions. In other words, you don't deal with LUNs and volumes anymore. You deal with VMs and decide what VM is important. And then all the storage and all the networking and everything happens behind the scenes automatically. So, how efficient those things are clearly becomes a point of differentiation. There's a wide amount of differentiation, but you will not be able to tell that without looking deeper.
What are the limitations of hyper-converged storage?
Taneja: All of these disparate technologies -- for example, data deduplication, compression, WAN optimization, backup, snapshot -- remember, most of these things grew up almost in separate companies. There were companies built on data deduplication alone. There are separate companies built on WAN optimization alone, and when you do a hyper-converged system, you're really bringing all those technologies in one box.
Clearly, each of these elements may or may not be best of breed in the traditional sense of the word. But the value of bringing all of them together in a hyper-converged format cannot be understated. A lot of customers I know would actually take [lower quality] individual elements from a best-of-breed perspective because the combination makes their life significantly easier than it is today.
For what types of organizations and what use cases does hyper-converged storage make the most sense?
Taneja: Hyper-converged systems are capable of taking on just about any application that exists in the environment. And in terms of the size of the organization, there is no particular reason why it is not usable by a small organization, a midsize organization or the largest of the large organizations.
However, since the primary purpose for hyper-converged systems is to manage a virtualized environment, those applications that are running on physical infrastructure would probably want to stay on the physical infrastructure. So, other than applications running in physical infrastructure today, I don't think there's a particular application that cannot be brought over on the hyper-converged infrastructure.