HCI vs. Traditional IT and Converged Infrastructure: Who Wins?
What are the differences between traditional IT infrastructure, Converged Infrastructure (CI), and Hyperconverged Infrastructure (HCI)? More than just a set of pros and cons, what are the fuzzy grey areas in between and overlapping these three approaches to IT infrastructure?
The Internet is littered with comparisons between traditional IT, CI, and HCI. They all have a bias. Someone is trying to sell you something, or they are trying to defend the infrastructure choices they are most familiar with. Terrible analogies are used. Buzzwords are watered down to meaninglessness, and hype rides atop it all.
Here at Hyperconverged.org, we have our viewpoint as well: as one might expect, we like HCI. But we're also honest enough with ourselves to know that HCI is a specifically shaped peg, and it won't fit into every hole. And while we use terrible analogies – and sometimes even buzzwords – it's worth the time to try to put these three infrastructure approaches in context, because the debate surrounding them is part of why HCI emerged as an approach to IT in the first place.
Before we can dive in to the differences between traditional IT, CI, and HCI, it would help to define what these three terms mean. Traditional IT is where each element of a data center's infrastructure is selected (mostly) independently of the rest.
The word "mostly" is in parentheses because traditional IT has never been quite so simply transactional. If one purchases servers from a given vendor, there's a decent chance that networking, storage, software, or some other component might also come from that vendor. Most organizations also make use of channel partners rather than buying directly from vendors, so the availability of a given solution from a channel partner is also a factor.
The advantage to traditional IT is that the freedom to pick any solution from any vendor means that an IT infrastructure can be designed to meet any need, and to fit any niche.
In addition, requirements dictate selection. If, for example, one required high availability to the extent that all network connections had failovers, including multipath for storage, then NFS-based NAS storage probably isn't on the table. It's been three years since VMware embraced NFS 4.1, but support for multipathing by NFS array vendors remains poor.
Similarly, organizations relying on channel partners will find the solutions under consideration narrowed to those solutions within the channel partner's comfort zone. Channel partners are engaged so that they may lend their expertise to product selection, implementation and/or ongoing support and maintenance. For decades this approach was not looked at as either particularly confining or particularly liberating; it was simply how IT was done.
The advantage to traditional IT is that the freedom to pick any solution from any vendor means that an IT infrastructure can be designed to meet any need, and to fit any niche. With traditional IT, the gating factor to finding the best solution for the job is the time available to research the myriad options available.
The flip side of traditional IT is that multiple vendors are involved, and organizations typically perform the final integration themselves. Having multiple vendors involved has led to "circular finger pointing" when an issue occurs, where vendors blame one another instead of taking ownership of the problem.
Upgrades are similarly fraught: the burden of planning for and then executing upgrades rests on the customer. Vendors may provide some limited support, but that support rarely extends to interactions of their product with the products provided by other vendors, unless vendor partnerships and mutual support agreements exist.
The existence of – and adherence to – product compatibility lists can help alleviate some of the support burdens associated with the multi-vendor nature of traditional IT. The cost of this, however, is again a limiting of the scope of solutions that can be considered for use.
This basic theme reiterated throughout the traditional IT section will be revisited in both the CI and HCI sections: the more one restricts the number of solutions to choose from, the easier that IT is to both support and maintain.
Long before CI was a talking point (or even an acronym), organizations turned to channel partners known as Systems Integrators (SIs). The job of the SI was to take multiple products – hardware and software – and make them work together. In some circumstances, SIs would even write custom software to marry all the various bits of IT together.
The important part was that SIs typically presented their offering to the customer as a single solution. The organization purchased "a payroll system," not a VMware cluster, some storage, networking, operating systems and application software.
The more one restricts the number of solutions to choose from, the easier that IT is to both support and maintain.
CI is basically systems integration: someone builds the solution for you, and sells it to you as a single SKU. Where CI differs from traditional systems integration is that CI aims to present a generic Infrastructure-as-a-Service (IaaS) to the customer, not a specific application or service.
Under the hood, CI often looks very much like traditional IT, often using the same products, from the same vendors, that would be used in a typical traditional IT infrastructure. CI is typically built using SAN or NAS storage arrays, though the use of HCI, scale out storage, or open data fabrics are possible as well.
The goal of CI is to allow customers to spin up VMs, containers, and even bare metal servers without having to worry about selecting, integrating or upgrading significant portions of the infrastructure. This is usually accomplished with a custom management interface, as well as with a combination of professional services that handle setup and upgrades.
The custom management interface included with most CI offerings reduces (but often does not fully eliminate) the need for administrators to use the management interfaces of the individual IT products that make up the CI solution. The custom management interface used by CI infrastructures can be thought of similarly to cloud computing, with one major difference: the defining characteristic of a cloud is the self-service capability of that cloud. CI vendors don't necessarily go that far, often being content with making resource provisioning "simple" for administrators, with “simple” being open to interpretation.
Compared to traditional IT, CI is quite limited in vendor selection. CI vendors have a very narrow list of partnerships. As a result, while CI can fit many niches and meet most needs, there are areas where CI is just not going to be the right pick, and either HCI or traditional IT will be called for.
Hyperconverged Infrastructure (HCI)
At its most basic, HCI is simply taking storage that is installed inside individual servers, combining it into a single, shared pool of storage, and then running workloads on those same servers. Essentially, it is high availability shared storage without the need for a SAN or NAS.
Hypothetically, HCI can work on any server, with any storage, in any configuration. In practice, however, HCI is the most restrictive of the lot. While some vendors offer HCI that can be installed wherever the customer wishes, the majority of the most popular HCI vendors either only sell pre-canned HCI appliances, or have a proscriptive product compatibility list.
HCI vendors can test the stack of hardware and software they offer more thoroughly than their competitors.
The result is that HCI solutions can be effectively thought of as being a fully integrated, CI-like, IaaS-in-a-can solution, provided by a single vendor. This provides a "single throat to choke" support experience that is quite different from the circle of blame that traditional IT frequently gets caught in.
With HCI, upgrades are also easier, in large part because the HCI vendor has a much narrower set of potential hardware and software combinations to worry about than either traditional IT or CI. HCI vendors can test the stack of hardware and software they offer more thoroughly than their competitors, and many HCI vendors are quite proud of how refined they've made the support process for customers.
Unfortunately, few things are easy to definitively classify in IT. A single vendor insisting on a narrow and proscriptive product compatibility list can dramatically restrict choices, even in traditional IT. Some flavors of HCI allow administrators to install the HCI software on almost anything, meaning that HCI can in some circumstances be more permissive than traditional IT.
CI, and especially HCI, have become popular in large part because they make working with IT infrastructure easier. While much of the key differentiation rests in how the various approaches to IT are sourced, purchased and supported, most CI and HCI offerings have a management experience that differs from traditional IT.
With traditional IT, administrators would have to administer each component in the data center. Every storage array, NAS, switch, server, and so forth would all have their own management interface. In many cases, mastering these interfaces became specialties, and even occupied entire careers.
CI solutions tend to offer a management interface that, at the very least, makes dealing with all hardware below the hypervisor less complicated. HCI solutions, on the other hand, can integrate hardware management, storage, networking, hypervisor control, data protection and even basic IT automation and orchestration into a single interface.
HCI is probably the least expensive, easiest to manage, and least difficult to maintain approach to most IT needs.
Today's IT teams are required to manage an increasing number of workloads with the same resources. In many organizations a significant part of IT spend is being devoted to reducing management overhead, especially among those companies using traditional IT infrastructure. The Infrastructure as Code (IaC) movement – i.e., writing code to manage configurations and automate provisioning of infrastructure in addition to deployments – can remove a great deal of the hassle involved in the set-up, upgrading and ongoing administration of traditional IT, but many of those same benefits can also be brought to CI and HCI.
IaC levels the ease of use playing field somewhat; however, IaC is a speciality in and of itself. This means that large organizations with well-staffed IT teams who embrace IaC may not see much direct management benefit from CI or HCI. They would mostly be searching for value in ease of upgrades, the ability to scale non-disruptively, or unified vendor support. For some organizations these are important considerations, while for others these are irrelevances.
The Right Tool for the Right Job
As much as we would like to be able to say definitively that HCI is the appropriate solution for all cases, this simply isn't true. HCI is probably the least expensive, easiest to manage, and least difficult to maintain approach to most IT needs, but there are always going to be niches where either traditional IT or CI are the better fit.
Very large organizations that want to push the limits of what's possible with IT will probably benefit from sticking to traditional IT, at least in part. Traditional IT, especially when combined with modern private cloud or container software, can operate at scales HCI providers can only dream of.
Similarly, specialty hardware may be needed to meet extreme uptime or availability requirements, to operate in extreme environments, or where extremely low latencies are required. Here, large organizations will most likely benefit from individually sourcing solutions, and keeping specialists for those solutions in-house.
Smaller enterprises, but which are themselves still quite large, are likely to benefit from CI. CI vendors will often be able to incorporate technologies that serve workloads at extremes beyond what HCI vendors can support, even if they’re not usually bleeding-edge.
Most organizations will see a benefit from using HCI, even if they do not use HCI for all of their IT needs.
In contrast, HCI can and does find a home with some of the smallest customers. Typically aimed at midmarket customers and smaller enterprises, HCI solutions are designed not to require specialists to operate, so they are a natural fit for smaller organizations.
In any case, most organizations will see a benefit from using HCI, even if they do not use HCI for all of their IT needs. HCI can meet most needs, and where it can be used, it should make the lives of administrators easier.