You are here

Hyperconverged infrastructure: Where to use it, how to choose it

public://pictures/ldp.jpg
Linda Dailey Paulson, Freelance writer, infrastructure, Linda Dailey Paulson Associates

Hyperconverged infrastructure is more than just one more buzzword among a confusing array of options associated with virtual infrastructure. It can lower overall infrastructure costs and reduce the complexity associated with adding new technologies. It can help your IT organization become more agile by combining the fundamental hardware and software needed to create a single integrated, responsive infrastructure that scales readily. And it can be a great first step toward building a software-defined data center.

Here's how it works, and where it might be a fit in your organization. 

Converged versus hyperconverged

A converged systems architecture combines compute, storage, and networking hardware resources within a single physical unit and includes a software control layer that enables the pooling of hardware resources and control of physical infrastructure through software. It uses commodity hardware and includes software-based management tools.   

Hyperconverged infrastructure is a variation of converged infrastructure. While the differences between converged and hyperconverged infrastructure are subtle, the key differentiator lies in how each system handles storage. All of the technology components in a converged infrastructure—storage, server, networking—can be thought of as building blocks that you can use independently. 

Although this suite of technologies came about as part of the larger “Let’s go virtual!” movement, hyperconverged infrastructure arose to solve a fundamental problem: the mismatch between virtual machines and data storage. IT needed to address that storage management complexity.

“From a definitional point of view, hyperconvergence refers to software that you can load onto industry-standard x86 servers, in turn turning those servers into a clustered, scalable pool of compute and storage,” says John Abbott, an analyst at 451 Research. 

Storage management was the fundamental driver behind hyperconverged infrastructure, says Rich Fichera, vice president and principal analyst at Forrester Research. But it also addressed “the rising complexity, cost, and operational ugliness of running these large environments.” That is, the complexity associated with management of storage, partitioning, virtual machine (VM) instances, wide area network (WAN) optimization, and other tasks. It also resolved some of the potential operational impact associated with data management.

The approach traces its origins to software-defined storage, which allows users to manage and provision their storage assets independently of the actual hardware. “The software-defined storage layer is the heart of the hyperconverged system,” says Fichera. It was born after users began clamoring for integration with an appliance.

The niche for hyperconverged infrastructure, Abbott says, has been “as an alternative to the traditional infrastructure components of separate servers and SAN-based storage." It also provides IT with a means of updating infrastructure—a need driven by more adoption by enterprises of technologies such as cloud, mobile, and social. "Hyperconverged software and systems are designed chiefly for virtual server environments and have been pitched as a simpler, more cost-effective alternative to traditional storage, such as storage area networks (SAN) and network-attached storage (NAS)," he adds. "This relatively new category has been driven chiefly by startups, and initial market adoption has been strong.”

Where hyperconverged infrastructure works best

Initially, most hyperconverged deployments focused on virtual desktop environments, says Abbott. "But adoption is now growing across a variety of virtualized workloads, including mission- and business-critical applications, as well as a range of emerging applications, such as Hadoop and data analytics.”

The technologies within this segment of the virtualization space, says Fichera, have matured to a point that they are perfectly capable of handling all but the most extreme enterprise workloads. "They are really worth looking at.”

A hyperconverged approach is not useful for all enterprise applications. It's not recommended, for example, for applications that require low latency and contend with high I/O rates. This includes, for example, architectures where you have 1 million input/output operations per second (IOPS) against a single set of files or tables or applications requiring less than a millisecond of latency, says Fichera. These types of requirements are, however, outliers.

Much of the interest around hyperconverged infrastructure has centered on regional and branch office applications, such as retail points of sale for a local store or an in-branch banking platform. Here, security is not an overwhelming concern, says Fichera, since all vendors can prove that they can protect and replicate data.

Moving to a hyperconverged infrastructure is chiefly about lowering total cost of ownership. Vendors can demonstrate that the approach has a lower cost per virtual machine over the lifetime of the system, Fischera says. But there's still much confusion in this space, including unwarranted worries about vendor lock-in.

Hypercoverged vendors: Startups versus the big boys

Several systems vendors, including Gridstore, Hewlett Packard Enterprise, Nimboxx, Nutanix, Pivot3, Scale Computing, and SimpliVity, offer hyperconverged infrastructure products. “Startups are leading the HCI charge, led today by well-funded and high-profile market pioneers Nutanix and SimpliVity,” Abbott says, adding that roughly $900 million in venture capital has been invested in hyperconverged infrastructure firms to date.

“Larger IT providers are slowly warming to the hyperconverged infrastructure opportunity, although to varying degrees of intensity,” Abbott adds. “Major market validation has come from VMware, with its Virtual SAN capability, while HPE, EMC, and Hitachi Data Systems have IP ownership in hyperconverged infrastructure. So far, Dell, Cisco, Lenovo, and NetApp have pursued a partner-led strategy around HCI. As such they can be viewed as potential acquirers in this fast-evolving market.” Things might change for Dell, however, due to its planned merger with EMC.

Customers can choose the way in which they implement a hyperconverged system between a purely packaged approach or pieces from different vendors, says Fischera. There are, he says, “substantive issues" about performance and price that are, in the end, “all about the cost per virtual machine.”

The four flavors of hyperconverged infrastructure

Vendors in this space, says Abbott, offer four basic flavors of hyperconverged products: hypervisor support; storage architecture; resiliency, which he defines as “replication or erasure coding as alternatives to RAID”; and form factor, which spans appliances, reference architecture offerings, and pure software models.

“Because there is some variation in the features, performance, and pricing offered by these vendors,” says Fichera, “users are responsible for knowing their own requirements, workloads, and composition of their storage.”

Users need not only to filter through the vendor offerings, but also to understand the changes to enterprise architecture and the ramifications of those changes beyond just storage. Potential users are ultimately comparing oranges, apples, and pears when shopping for hyperconverged infrastructure products. It’s not a simple comparison because of the many different ways in which these types of systems are bundled.

“Vendor lock-in is a distraction. … Under the covers it gets complex,” says Fichera. Choices can range from “VSAN on HPE servers—pick and choose the server and the software-defined storage option or go with one whole vendor for the stack. There are a lot of ways to cut up this elephant.”

A few caveats

To save costs, enterprise IT may be tempted to adopt a “roll-your-own” approach to hyperconverged infrastructure. Fichera cautions against it. “It’s a unique set of ingredients in the secret sauce that is hard for internal development to do,” he says.

In that same vein, some IT organizations may face operational challenges in adopting hyperconverged infrastructure, since the storage and virtual ops teams may not want to give up their turf. Storage commonly loses jurisdiction, says Fichera. But the blurring of storage and server networking "turns into one of the pros, which is the cost of administration.” IT organizations save money in the aggregate, he says.

Part of the challenge for those interested in hyperconverged infrastructure is that there is currently “a lot of posturing” among vendors, Fichera says. Some vendors are making money selling richly configured systems. Others have discovered that, instead of selling enterprise-grade servers—a market that has been flat as of late—they can reposition as a provider of hyperscale rack servers.

As Fischer points out, the margins on selling hyperconverged systems to enterprises are much higher than for selling them to cloud providers. One sweet spot to which vendors are selling is the second-tier cloud services providers—the regional cloud service provider or telco interested in adding value-added disaster recovery/backup to its services menu, which are buying hyperconverged infrastructure for these offerings.

Is hyperconverged for you?

Organizations should adopt hyperconverged infrastructure to save money and reduce the complexity associated with network management and operations across servers and storage. For this purpose, analysts say, IT organizations of all sizes are finding that the technology meets their needs for a wide range of mission- and business-critical applications.

The way to proceed is to start with a small project and then scale up. Start by surveying your existing infrastructure, and know what you want to achieve by moving to hyperconverged infrastructure. This will help you not only to refine your needs, but also to identify the features you truly need so you can focus in on the right vendor offerings.

Image credit: Flickr