This content is part of the Buyer's Guide: Find the HCI appliance that meets your storage needs

How to find the right HCI appliance

Choosing a hyper-converged appliance can be a difficult road to travel. Expert Ed Tittel provides the road map for selecting the right hyper-converged tool for your business.

When it comes to choosing a vendor product for an HCI appliance, there are dozens of options available, and more are coming to market frequently. All of these options make selecting the hyper-converged product that best fits your company a daunting challenge.

In the sections that follow, the top hyper-converged vendors' offerings will be assessed against a set of seven criteria, which combine to produce the overall evaluations presented in the conclusion.

The purpose of a hyper-converged appliance is to make it easy to deploy and operate a compute and storage architecture. Some of the products covered here are simple building blocks with limited scale. Other products support serious to extreme scaling, while some are specialized for specific use cases including database access, virtual desktop infrastructure (VDI), server consolidation and more.

The majority of HCI appliance vendors offer their own appliances or carefully vetted ones that combine compute, storage and networking capabilities. Hewlett Packard Enterprise (HPE) is the exception, packaging its own software on HPE-built servers. Most of the vendors discussed here support small and medium-sized businesses all the way through the enterprise.

Simple and complex HCI appliances

Each of the vendors' offerings requires some IT skill and knowledge to understand and navigate. The amount of knowledge necessary is assessed in terms of how simple or straightforward it is to match workloads to products.

VMware makes the argument that its offerings are simplest because it controls the hypervisor and it provides a tightly coupled software layer for storage. But would-be buyers have to pick hardware from a hardware compatibility list, for which they may have to hire an integrator (or the hardware creator) to help pull systems together, or they have to choose from a list of prefabricated options.

HPE makes the argument that its offerings are simplest because it provides its own hardware and it has tight relationships with hypervisor providers VMware and Microsoft. Yet HPE offers an alarmingly large number of building blocks and options from which systems must be composed, with costs varying widely. HPE's offerings cover a complex array of options, but they do their best to guide customers toward a best fit.

SMB-centric Scale Computing offers a simple collection of elements and options that are truly easy to mix and match as needed.

The rest of the HCI appliance vendors fall into the middle of the spectrum. SimpliVity, Stratoscale and Pivot3 fall closer to the simple side of this continuum, while Maxta and HyperGrid (formerly Gridstore) -- along with VMware, HPE and Nutanix -- fall more on the complex side.

Clarity and intelligibility in a hyper-converged appliance

Digging into each vendor's products requires working through varying amounts of material and information to figure out the right size and performance that you need. Understanding what needs to be purchased and how much digging must be done to arrive at that point is probably the steepest climb on the hyper-converged appliance learning curve.

By virtue of its extreme simplicity and fewer options, Scale Computing stands alone when it comes to providing information for evaluating its hyper-converged appliances. SimpliVity, Stratoscale, HyperGrid, Maxta and Pivot3 are all clustered closer to the "clear" and "intelligible" ends of this continuum. HPE comes next, and falls a bit short only because of the amount of information and material needed to support the large array of options it offers to its customers. For what the company offers, HPE does a good job. Nutanix and VMware bring up the rear because of the large number of choices buyers must make when determining what they want and the large amount of material to read, absorb and ponder before they can do so.

Building blocks and costs

Some HCI appliances are composed of a small number of models with straightforward pricing. Others present a sizable number of options, including selections for compute and storage. In some situations, the only way to determine cost is to use a vendor-supplied calculator.

Many vendors won't discuss pricing unless presented with a specific workload for them to model, and even then, they will only discuss it with that customer. A common policy is to talk pricing with customers only when they've decided what they need to buy.

Here, again, Scale Computing stands by itself in terms of providing clear, straightforward descriptions (and costs) for building blocks that users put together to create a complete system. At around $25,000, Scale Computing's entry-level offering appears to be significantly less expensive than that of any other major competitors. SimpliVity and Pivot3 come next -- a ballpark figure for entry-level systems from each vendor is in the $50,000 to $60,000 range. The rest of the HCI appliance vendors fall somewhere in the range of $60,000 to $100,000 for entry-level offerings.

The majority of HCI appliance vendors offer their own or carefully vetted appliances that combine compute, storage and networking capabilities.

HPE customers must sort through a large number of options. There are also many choices with VMware's Virtual SAN, as buyers must select from a large array of prefabricated configurations with software pre-installed or from select hardware elements on a compatibility/recommended SKU list. The remaining vendors fall in the middle of this range, with SimpliVity, Pivot3 and Maxta on the lower end of the continuum, offering a smaller and more manageable collection of building blocks. Nutanix, HyperGrid and Stratoscale shade from the higher to the lower end of that same continuum because of their larger collections of more complicated building blocks.

Overhead and efficiency

Some hyper-converged products are architected to maximize performance for specific hardware and workloads; other offerings are architected to support the widest variety of hardware. There will be tradeoffs involved with either type of offering, so it's important to perform proof-of-concept implementations that model prospective workloads. This is really the only way to understand what you'll be buying (and getting).

Several vendors make much of providing their own hypervisor based on the open source Kernel-based Virtual Machine (KVM). Having a KVM means customers don't have to purchase a VMware hypervisor license. Nutanix, which started out in 2011 as a strong VMware partner, now has its own KVM-based product, Acropolis Hypervisor. Still, the majority of Nutanix's customers still use VMware hypervisors.

All hyper-converged storage vendors prefer their block-level I/O being mediated through the hypervisor, rather than relying on virtual file systems of any kind. These vendors, including Scale Computing, Stratoscale, SimpliVity (in some implementations) and Pivot3, claim to outperform systems based on either VMware or Hyper-V, sometimes by significant margins.

Other vendors, including HPE, Stratoscale, Maxta and HyperGrid, are hypervisor agnostic, and make claims about optimized IO based on their efforts to work through their own purpose-built VMware or Hyper-V drivers. VMware also claims to offer optimization for all elements of the hyper-convergence software stack (hypervisor, storage access, OS and application).

VMware is a major player in the hyper-convergence market, mainly because it owns more than 70% of the hypervisor market and a large portion of the storage services market that runs alongside it. As such, understanding how a competing vendor positions itself against VMware is an important measure of performance for other players. You will either see other vendors address this directly in datasheets and benchmark results or can ask them to provide such information. This is of great interest and value to companies seeking to understand if VMware is the best provider for their hyper-converged products or not. Those companies with existing or substantial VMware investments will probably be most inclined to add VMware's hyper-convergence products to their technology arsenal. However, hyper-converged vendors who allow customers to bypass VMware hypervisors claim to eliminate the "hypervisor tax" included in VMware licensing.

Hyper-converged vendor knowledge and support capabilities

For support of a hyper-converged product, you have to consider the hardware supplier, the hypervisor creator, the guest OS creator and at least one application or service vendor (such as a typical VDI or server consolidation effort or support for a database management system (DBMS), big data platform, web and application server and so on).

Hyper-converged products differ in terms of the number of actual parties involved. They also differ in the terms and conditions (and types) of support that they make available; any or all of these must figure directly into your choice. All of the vendors claim to be able to deliver complete support without having to rope in other third parties to help solve problems or troubleshoot.

Because they offer customer testimonials and case studies to demonstrate their all-encompassing support coverage, the claims of Scale Computing, Pivot3 and HPE appear to be the most credible. Next in line comes SimpliVity, Stratoscale, HyperGrid and Maxta, in no particular order. After that, VMware and Nutanix finish out the pack.

Be sure to understand each vendor's support offerings and terms as you work through your shortlist of candidates during your selection process. Over time, the cost of support can eclipse that of the hardware, especially when you have to pay different vendors for hardware and software support, as is often the case.

Special features and capabilities

There are lots of little differences, wrinkles and extra options available in HCI appliances. Some vendors offer custom configurations for specific workloads, such as big data, specific database engines and more. Other vendors support hardware compatibility regimes, where hardware vendors can submit their components and systems to be vetted for suitability. These will be tested to make sure they can run specific hyper-convergence software. In some cases, such vendors may even offer ready-to-configure systems that include pre-installed software and base-level configurations, which need only be supplied with local parameters, IP addresses and so forth to be made fully operational. All of the vendors offer the ability to add compute capability or storage capacity.

Some vendors offer environments for specific workloads or usage scenarios. HPE leads in this area with a wide variety of specialized offerings for big data, various DBMSs and other specific workloads. But because understanding workloads is an important key to delivering a suitable hyper-converged appliance to match, this is something all of the vendors must address as they work with buyers to negotiate a purchase.

The largest scale implementations currently fall to HPE, VMware and Stratoscale. Most vendors should have no trouble populating even the largest data centers with HPE, VMware or Stratoscale products, although Scale Computing operates at the lower end of this continuum because -- except for certain VDI scenarios -- their maximum cluster size tops out at eight appliances.

Next Steps

Hyper-converged infrastructure presents unique challenges.

The number one concern for hyper-convergence appliance vendors? vendor lock-in.

Read about the hyper-consolidated IT infrastructure.

Dig Deeper on Converged Infrastructure Management