Hyperconverged Infrastructure

How Hyperconverged Infrastructure Is (And Isn't) Like a Car


One of the key underpinnings of ease of use is standardization. Computers are often compared to cars, with cars standing in as a sort of generally accepted standard of ease of use for a mass-market mechanical device. So, are computers like cars yet? And if so, how does hyperconvergence help or hinder the journey toward this highly subjective measure of ease of use?

If cars are the level of ease of use we're aiming for, then we need to consider why cars are considered easy to use. One frequently cited element of a car's ease of use is the universality of the driving experience, so it’s worth taking a look at the concept of standardization as it applies to both computers and cars.

Clearly, not all cars are the same; differentiation between vehicles is important for driving sales. The physical design, engine power, luxury features, and more will vary from model to model. This creates very real differences between cars. One car might great for someone who’s 6-foot-5-inches tall, but another one over there might be undrivable by the same individual.

At the core of most cars, however, is a common user interface with two main branches: manual or automatic transmission. A steering wheel turns the car. There’s a rear-view mirror and side mirrors. There are controls to the sides of the steering wheel for turning signals, windshield wipers and other common controls.

Anyone who’s driven a car before – and who can physically fit into the car to drive it – should be able to drive almost any car off the dealer's lot with very minimal training. This is the power of standardization.

Standardization in IT

Standardization is something of a fraught subject in IT. Most standards for computers emerge only years after novel functionality is first created. When they do emerge, they rarely emerge alone: it isn't abnormal for there to be half a dozen different standards for solving the same problem.

The universality of "turn the big wheel left to go left, and right to go right" isn't really a concept that IT vendors work toward. IT vendors, like most vendors in the majority of markets, want to create "friction" – also known as lock-in – which keeps customers within their ecosystem. IT vendors want customers to feel that migrating to a rival's offering will be more troublesome and expensive than any supposed cost savings one might find with another vendor.

This is an approach to business that doesn't really work for cars. The gas pedal, for example, must be on the right, with the brake on the left. That's the law.

In the case of pedals, the most obvious reason for legally mandating vehicle interface standards is so that, when under pressure, a person's muscle memory from a previous car translates to the new one. Nobody wants drivers hitting the gas instead of the brake when trying to avoid an accident because two different car manufacturers decided to differentiate their vehicles by changing up the pedal positions.

These sorts of laws don't really exist in IT.

Let's also consider the process of refueling a car. Laws exist which define the physical characteristics of fuel dispenser nozzles and receptacles. These laws exist so that cars can refuel at any refueling station that carries their type of fuel, and in some cases to prevent consumers from putting the wrong kind of fuel into their vehicle (though a lot more work could have gone into the latter).

If car companies were free of legal constrains to operate as IT vendors do, they would own the fuel distribution networks, and would have made proprietary fuel nozzles and receptacles so that customers who purchase their vehicles could only refill at fuel stations owned by the vendor. They would patent both the nozzle and receptacle, and legally attack anyone who tried to make an adapter to allow a car to fuel up at a competitor's fuel station, or anyone who tried to create a multi-vendor nozzle system.

HCI and Standardization

Hyper-converged infrastructure (HCI) exists at least in part to the above silliness. This isn't to say that HCI is particularly standardized: one vendor's HCI solution is entirely incompatible with the next, and looks to stay that way for the foreseeable future.

HCI is a means of lowering the cost of storage by bypassing the traditional storage technologies, and the vendors that fueled them. Storage vendors have made serious effort in the past to lock customers in to their solutions. As a result of the high friction of changing vendors, storage vendors as a whole could – and did – keep storage prices high.

This created a market gap for HCI vendors. Storage costs were high, nothing was particularly interoperable, and shared storage of the sort needed for increasingly popular virtualization solutions also required expensive networking dedicated just to storage.

HCI vendors incorporated the storage into the compute servers, and interconnected the storage between the servers using standard Ethernet networking. In doing so, they could drive the cost of storage down. This idea proved sound, and a number of different vendors took a swing, each with their own proprietary twist.

With the storage stranglehold broken, prices were kept low. The lack of lock-in and thin margins forced vendors to compete on merit, and this in turn has driven a renewed interest in creating user interfaces that are intuitive, discoverable, and otherwise easy to use.

Like most things in IT, HCI standardization is a mixed bag. Solutions are not particularly standardized, at least at the storage level. HCI is typically used as the underlying physical infrastructure for a virtualization solution, however, and those are standardized.

HCI customers can take their workloads from one virtualization platform to the next with a minimum of fuss, even if they can't mix and match nodes from different HCI vendors to build a heterogenous HCI cluster. On the other hand, HCI helped drive the cost of the physical infrastructure low enough that for many organizations entire clusters can be treated as essentially disposable: one can usually get storage plus compute today for less than storage alone would have cost a few years ago.

Computers aren’t as easy to use as cars, and lock-in still persists in many places. But in between peaks of expensive, proprietary shenanigans there are valleys of fierce competition.

HCI has created exactly such a shift in IT. Now, and for the next several years, HCI will keep the cost of the underlying infrastructure low enough that administrators can focus on the usability, portability and interoperability of the workloads for which they’re responsible, rather than fretting about the price or standardization of the underlying infrastructure.