This post originally appeared in Human Infrastructure Magazine, a twice-monthly newsletter on being a human in IT. You can read back issues and sign up for free, and we don’t share your information with anyone.
In the SDN world, there are some customers who insist on completely “open” technology. That is, they want whatever technology they choose to have a shot at working with whatever else they choose, because all of the tech is open source. Vendor lock-in is anathema to these customers.
The reality of SDN products is that, despite the cry of open-everything, the most useful solutions are vendor-specific. The solutions might or might not have open source components, but that doesn’t especially matter to the customer who is simply trying to buy a turnkey solution that works.
Perhaps “turnkey” is the crux of the matter. Most organizations can only make use of a tool that actually does something. They have neither the time nor expertise to assemble Lego blocks into something useful. They need to open the proverbial box, plug it in, and experience success.
In this context, we see the uptake of Cisco ACI, VMware NSX, and SD-WAN solutions such as CloudGenix and Viptela. These tools perform different functions under the broad umbrella of SDN, but make a special point of emphasizing out-of-the-box usefulness—with no assembly required.
That’s not to say you don’t have to configure these platforms to do something. Certainly, you do. You might even want an integrator to help you get it right. But you don’t have to write your own software to stack the Lego bricks together.
Idealism Vs. Usability
The open-everything vs. turnkey mentality is one I’ve had to come to grips with as I’ve watched the networking industry change. As end users, we all think we want interoperability, openness, and the chance to swap out any individual pieces and parts whenever we want to. Only, that’s not actually true. That’s idealism talking.
Idealism aside, I believe most people simply want their IT to work with a minimum of fuss. Does that mean vendor lock-in? At this point, probably. Why? Because IT infrastructure is increasingly interdependent, creating ever more complexity.
With the software defined data center, we’re taking formerly disparate tech and mashing it together into a unified whole. Systems talk to one another through API calls. Central orchestration software, driven by policy, creates changes in compute, storage, networking, and security. This makes for a fragile operating environment.
Attending an OpenStack Summit Boston 2017 keynote, it was clear that major OpenStack directives going forward are maturity, stability and usability. There’s less emphasis on new features, and more emphasis on polishing what’s there.
Attending a session, I saw a frightening OpenStack reference architecture. Imagine Kubernetes (K8s) as the base layer of OpenStack—the foundation that keeps OpenStack running.
And then imagine tenants running their own K8s or Mesosphere on top of THAT to create their own multi-tenant environments. Admittedly, this architecture is only interesting to a small number of cloud operators. Most operators wouldn’t need the top layer that completes the OpenStack sandwich. But still…is OpenStack running on K8s a private cloud stability improvement?
What enterprise has the energy to expend on an orchestration service so fragile that it takes yet another orchestration service to keep it running? Every instinct in me as an IT architect screams, “Run away!” when contemplating such a design.
If we, as an industry, can’t run OpenStack reliably on bare metal, then I think it’s game over. We’ll never see mass adoption. Although, in fairness, that was sort of the point I referred to from the keynote earlier. It’s time for OpenStack to grow up.
That lengthy illustration returns my thinking to turnkey solutions suitable for consumption by the average enterprise. Suddenly, turnkey doesn’t sound so bad, does it? Hmm.
Well, what about the no-no of vendor lock-in? At this point, I argue, “Who cares?” If you’re buying on a 3-year cycle, you’ll be tossing out what you’ve got in the not-too-distant future anyway. Therefore, go ahead. Get locked in if what you’re buying is easy to use and solves your business problems.
And by “solves your business problems,” I mean…
- Enables your product or service to come to market quickly and remain competitive across iterations.
- Enables excellent customer service such that you retain customers, and grow or at least maintain revenue.
- Meets your budget requirements, with Opex as critical to analyze as Capex.
Many of you will benefit from partnering with the right combination of turnkey vendors to take on the technology integration work for you. If they did a good job with both their product and integration with other products, you’ll be well-positioned to make SDN work for you.