We Need to Move from Assembling Car Parts to Driving Cars

During a great conversation I had with Terry Slattery during Interop New York, he said “well, I don’t think anyone should be configuring VLANs and asking ‘How to configure a VLAN on a switch’ – we should be focused on providing end-to-end connectivity”, and there’s absolutely nothing in that statement that one could disagree with.

However, someone will eventually have to know how to troubleshoot VLANs, and when presented with that argument Terry said “Yes, that’s absolutely true – but that would be like a car mechanic looking at the engine in your car”, and then it struck me: one of the real problems we have with our networks today is that networking industry is a car parts industry. Instead of focusing on how to build and sell marvelous cars, the whole industry is busy selling car parts to home builders, who use them in a hodgepodge manner to get their unique contraptions.

Why would that be the case? Every time a new entrant is trying to do a disruptive move (and it could be 3Com or Cisco in the 90’s, or HP a few years ago when they decided to enter data center networking), they tell the customer that they can replace the parts from the existing “car” with a cheaper, newer, shinier and mostly compatible parts made by them, which inevitably focuses the discussion on specifications of the “car parts” instead of usability of the car and the ease of driving around (which should be the whole point of owning a car for most people).

To change that, we need a fundamental shift in behavior, where we start buying solutions (cars) instead of parts (routers, switches, firewalls…) that we carefully glue together… but of course that will never happen because that would be the ultimate lock-in worth millions of dollars and nobody is willing to bet their future or the future of the whole company on a single-vendor car (unless, of course, that vendor is IBM or Oracle).


  1. I thought the debate was that "networking didn't change much over the years" it was that simple end to end or completely purpose built "car". But virtualization was the disruptor in all the other purpose built areas, compute, storage etc. Thus, networking had to catch up or beef up to a truck to carry the new load. That was the main issue relating to the VLAN analogy given so the network industry went with the, buy a whole new purpose built truck UCS with fabric type of platform or "bolt on" approach with overlays(nitros and big tires) etc to accommodate all the other pieces to complete the virtulaized car.
  2. The analogy is good. I think some small and medium businesses don't want to own data centers and infrastructure... they either go public cloud or have a managed private cloud. That's buying the whole car, a whole solution. The guesswork is removed but you may void your warranty if you want to customize it too much.

    Larger business... I think it's more like racing than freeway driving. You want a pit crew and you want to be able to micro tune based on your needs. A total solution will not provide the flexibility you need. Not technologically and definitely not cost-wise.

    Business needs change, vendors innovate, competitors cause the market to shift much more rapidly in the past. Rather than buy a total solution, I think there is more value in buying layers or sections of infrastructure. In other words... explore all potential orchestration vendors out there and pick the one that integrates well with your existing and planned infrastructure (external, internal, hybrid). Likewise, pick you hardware, pick your hypervisor, pick your network layer vendors. There's no 100% guarantee, but good selections and good design should enable you to swap vendors at any tier more or less painlessly. If you get a smoking deal on a different server brand, you can phase that into your overlay and orchestration. Some new vendor comes out with a converged solution with appealing price points? You should be able to take advantage of this opportunity without impacting what you already use or being forced to buy separate management, orchestration and tools. Same goes for the Orchestration layer... If vendor Z has new, amazing widgets and knobs, you should be able to stand up that suite and migrate some underlying infrastructure to it seamlessly.

    Yes, I know there will never be the nirvana of pure Open, pure compatibility... but I think it can be done much better in layers than any one vendor's end-to-end solution.

    Just my opinion.
Add comment