The shift from the mainframe to modern epochs in IT infrastructure has been well documented numerous times. Depending on the particular storyteller, the tale starts with deeply-integrated, vertical stacks that encompass everything from compute to networking in large systems that required small armies of people to deploy and manage.

As technology matured and the Internet exploded, what were once integrated functions were broken out into standalone components. This led to the second real epoch in IT infrastructure, most notable because it gave birth to the best of breeders.

This second epoch has served us well. We saw the birth of new architectures. We have perched upon new performance peaks. We have mastered new capabilities that have transformed the whole of IT. And because we are cobbling together our solutions from standalone components, we have been able to do all of this while we customize the architecture to suit our sometimes very unique needs.

But all of that has come at a cost: namely, complexity. We might experience that complexity firsthand in the form of policy madness. Or we might be secondhand consumers, choosing to deal with it by paying for service integration. Whichever we choose, we ultimately pay in one way or another.

The costs of infrastructure assembly—both time and money—are already shaping the market. Efforts like VCE and Exadata are good indicators of a change in how infrastructure and applications are consumed. Rather than relying on specialists who integrate products upon delivery, some customers are choosing to purchase already-integrated solutions.

The distinction here is integration as a service vs. integration as part of a product. In the former, integration is the outcome of some effort that happens after a product is shipped. In the latter, integration is an actual part of the product. The difference between the two might seem subtle, but it is important in terms of how the market will evolve.

If you believe that buying patterns will shift more towards a re-verticalization of IT, then individual constituents in the current IT stack will ultimately align to a relatively small number of overarching IT providers. You won’t have an integration layer in the market that handles this function (or more precisely, that layer becomes much smaller).

Today there is a swath of resellers and systems integrators that perform this integration function. In a re-verticalized market, the large suppliers handle the integration burden by assembling products that are already integrated and providing both the services and support around the final infrastructure as a whole.

This makes perfect business sense because right now those large players are leaving money on the table. If they can act as their own integrators, they can keep pricing low (essentially lowering prices but maintaining margins by cutting out the supplier in the middle).

Of course, this is not a transition that everyone can make. The largest suppliers have the customer relationships to pull this off. The smaller suppliers do not have the product center of gravity, and they lack the customer reach to operate at the kind of scale necessary to dominate.

These dynamics form the major case behind the mass consolidation of infrastructure suppliers, and arguably this is why Cisco’s CEO John Chambers has both set Cisco’s targets on being the number 1 IT supplier, and predicted a brutal consolidation of the networking space.

It is conceivable that the market can only support a small number of very large companies. Oracle and Cisco would seem poised to do well here, as would companies like HP. Dell is interesting depending on what other steps they take from a solutions perspective. IBM and maybe even companies like Intel start to look interesting.

There will be very large companies that end up on the outside looking in. It would not be surprising to see multi-billion-dollar companies struggle to grow in a changing environment like this. They might not outright die (epochs in IT are measured in decades, not individual years), but they could see a radical tapering of growth, which typically leads to other issues (talent, product pipeline, and so on).

Now while this new epoch of IT might be characterized by the mega suppliers, it won’t be only the large players that have a role. It is possible to deliver components as part of an integrated solution. But the platforms themselves will be fundamentally different, which means the next epoch components must be designed with a different endgame in mind.

The question is really about how companies exist in the second epoch while simultaneously preparing for the third. This is nontrivial – enough so that it deserves its own (future) blog post.

[Today’s fun fact: The first lighthouse to use electricity was the Statue of Liberty. Come on, baby, light my fire.]
Showing 4 comments
  • David Crosbie
    Reply

    Hi Mike

    I think there is another aspect to this story:

    You are seeing different organizations on radically different efficiency curves – essentially the ones (Google, Facebook, Amazon) that have turned their back on the vendors and are building these vertical stacks themselves have a solution that is 10x better (cost, performance, flexibility) than those players (Target comes to mind) who are totally dependent on 3rd parties.

    If Cisco et al does not change models then rational IT departments are going to outsource their operations lock, stock and barrel to the 10x players – and that will shut Cisco out because the 10x players roll their own hardware.

    What this really means is that Cisco has no choice but to ride the SDN wave as fast and as hard as they can – not for the cost perspective but because it enables them to make this more to truly flexible stacks.

    Verizon and the other outsources suppliers need to get on the same efficiency curve.

    Vendors like NetYCE (which I know and like) who are pushing a sort of SAP for IT departments have the other part of the secret sauce – the ability to organize the design and deployments. In essence bringing the Kanban system from the automotive into the networking industry.

    Cheers

    David

    • mike.bushong
      Reply

      It’s hard to make the case that the webscale guys who are building their own have a huge cost advantage because you need to account for the carrying costs of maintaining that system. I am not saying they don’t have an advantage (depending on scale, they might certainly have that advantage), but we should not just measure the cost of device here.

      I think the logic arrives at the same place though: that there is a shift in how people buy that is here, imminent, or not that far off (depending on your frame of reference). I doubt the webscale guys end up selling infrastructure, but that they could is interesting enough to create a shift in market dynamics. The business models are different, the selling motions are nontrivial, and the support side is much more intense. But you’re absolutely right that this puts pressure on everyone else.

      Would love to see you expound on the role of SDN that you mention.

      -Mike

  • David Crosbie
    Reply

    Now this is where I disagree with you Mike.

    I think we are seeing the start of a radical divergence in efficiency in data centers akin to the divergence between the US car manufacturers and Toyota in the 1980s.

    In that case Toyota evolved a radically different approach to the production line (the US kept the line running at a constant speed irrespective of what came out the end, while Toyota empowered anyone on the line to stop the entire until a problem was fixed).

    You can see this today in every data center I go to. A network engineer finds that there is a problem with the configuration of a box. He/she fixes that problem by logging into the box command line, makes the fix, and then goes onto the next problem. In the Toyota system he/she would fix the problem, document it and then launch a process to both determine why it was misconfigured (one off or systematic) and which other boxes were affected – and then correct them before they caused a problem.

    So I stick with my prediction that organizations that are on the path of radically re-imagining their IT operations will in five years end up in a radically different place than organizations that do not. Those that do not evolve significantly in the next five years will then have no choice but to outsource their operations to a third party.

    • mike.bushong
      Reply

      I don’t disagree with anything you wrote. I am not actually sure where we disagree to be honest. I think I just don’t see Google competing with Cisco in a classical sense, but your comments here make sense to me.

Leave a Comment