Cloudmageddon: the bifurcation of builders and buyers
Apr 17, 2017
For decades, the people who bought infrastructure were the same people (or at least working for the same company) as the people who used that infrastructure. Put less obtusely, companies were responsible for buying, building, and ultimately using the compute, storage, networking, and applications on which their businesses relied.
That dynamic is changing, and I believe there will ultimately be a bifurcation of the infrastructure markets. This will fundamentally change how this technology is bought and used, and likely how it is developed to suit those markets.
Builders vs. Buyers
To start, let me group those involved in this market into two high-level buckets: those who build infrastructure and those who use it. The former is typically the IT teams, and the latter is the rest of the company who complain about the IT teams.
On the user side, imagine a spectrum of different types of users. Those closest to being builders use a DIY model where they design the architecture, buy the equipment, configure and operate the network. If you are a little less skilled or require slightly less control, you might adopt reference architectures, which essentially reduce some of the complexity of designing and building from scratch. Solutions are prepackaged and tested (I include things like hyper converged infrastructure here). And of course cloud is the ultimate in un-owned infrastructure.
For builders, the most basic builders treat their infrastructure as a collection of devices. For more control, the builders might be more concerned with the components (individual line cards or perhaps a specific interface). And even more sophisticated builders will treat anything on the device that has some state associated with it as an individual element to be polled or controlled.
Evolution of the distribution
The distribution across these types of builders and users has been clustered at the center of the spectrum. Typically, smaller companies with less staff and fewer applications to manage live at the outer edges of the user spectrum. And large cloud companies with tons of staff and an appetite for differentiation through architecture live at the outer edge of the builder spectrum.
But the cloudification of IT (or cloudmageddon if you are on that side of the debate) is likely to change the distribution.
The industry will ultimately end up with a barbell distribution where users migrate to cloud and managed cloud, and the builders are predominantly building responsive cloud environments where infrastructure is a disaggregated set of subsystems that provide very granular control.
Users and convenience
For users, the trend is clear: infrastructure generally (and networking in particular) should more closely approximate the power grid. When you walk into a conference room, for instance, you don’t even turn lights on anymore; lights just happen. Similarly, when you want to use an application, the network is the last thing you want to be thinking about.
Before network engineers mistake this for saying that jobs are going away, anyone who manages power grids will tell you that a lot of design and expertise goes into making the power grid work seamlessly with the surrounding infrastructure. Similarly, making a network work as an always-on, never-cared-about resource is not an easy task.
But unless infrastructure drives a sustainable differentiating advantage to the rest of the company, an increasing number of users will opt for the convenience of having someone do it for them—importantly, even if it costs more.
Builders and control
Builders have a different math. Builders need control so that they can build fit-for-purpose environments. In this case, purpose might mean that the infrastructure is optimized for cost (think: early cloud), applications (think: what Oracle will do), compliance (think: data sovereignty), or even user control.
That control will come through disaggregating the elements of the infrastructure. We see this with white box trends, but the real driver here will be all things dynamic. Anything that throws off some state will be treated as an individual element. This should drive vendors to move from device state to subsystem state, and it will drive the need for streaming state, state collectors, and ultimately some way of reconciling information across the different state producers so that some action can be taken. This is likely where a lot of the machine learning hype will have to eventually go, but this is not a machine learning post.
A change in buyer
The real change here is that where the users and buyers were one and the same, in this new distribution, the builders become the buying proxy for most of the users. This isn’t actually that prophetic; we already know that cloud properties buy a lot of stuff. But if the trend continues through cloud to managed cloud, it does mean that there will be additional opportunity.
First, there will be an even more rapid rise in companies looking to do the managed part of managed cloud. These companies will put pressure on and eventually displace some of the VARs whose primary role is to handle integration in the current world order.
Second, if this managed business is big, it will take a strength of the large telcos and pit it against the building strength of the cloud providers. The cloud providers will need to develop managed service arms along with more traditional enterprise sales teams. I believe this means that the large telcos are not as doomed as is sometimes fashionable to predict. While some may stumble (and even die), I think the larger ones will survive, albeit with an evolving business model.
Third, the more distant from the infrastructure, the less important the brand to the end user. This should help the challenger brands to some extent and it loosens the shackles of incumbency. But it complicates things for vendors as they must split their marketing and sales motions to address both the builders who buy and the users who consume. The managed service providers (both telco and CSP) will only pay for product that is consumed, and they will expect their vendors to drive brand preference if they want their unfair share of the pie (else purchases will default to standard offerings).
The bottom line
Strategically, straddle positions (believing that everything will be a little bit of A and a little bit of B) is probably the most comfortable position to be in, because it means you don’t actually have to make a decision. This is true, of course, until the straddle makes you ill-equipped for either A or B. Companies will ultimately have to bet on what the final outcome will be: do they believe buying power resides with the builders and users. And that decision will have profound impacts on product strategy.
Just as important, the people whose jobs are based on this dynamic will also need to evolve. While the transition is not going to complete any time in the near future, those who plan to stay in this arena long-term should develop a point of view. As with most transitions, those who move the right way first tend to reap a disproportionate share of the rewards.