7-10 years for new infrastructure layers to win
7-10 years for new infrastructure layers to win
With VMware going all in on kubernetes, my co-host Brandon says the next phase for this market is offering the best way to run things on-top of kubernetes. This is a shift from running kubernetes itself, and relying on a toolchain and layers to package up app code into containers, do all the network and prod confirmation, security enforcement, and deploying it all.
You could position this as climbing up the stack from IaaS to PaaS. But, if his theory is right, it also means it's the tipping point for the IaaS/cloud native layer, that kubernetes is the winner.
I have a vested (pun!) interest in VMware and Pivotal doing well here, but never mind that.
Instead, I want to point out how long this took, from the original release of kubernetes 5 years ago. This will give us a rule of thumb for timing how long adoption for infrastructure tech takes an even a hint to enterprises for when they should seriously consider using the tech.
Five years, then, is a reasonable time to start with. For example, if you were a startup in this space, your clock to an exit would be around five years. If you were an existing company, your planning and finance business case would be about five years. That is, companies working on infrastructure software can set their roadmap expectations to around five years.
This also suggests how long conservative (or "pragmatic," to use optimistic diction) users of that infrastructure should wait until they adopt it. It's unreasonable to think users will start the clock at the creation of project, so you could peg their clock at two to three years from the birth of the new infrastructure layer.
Anecdotally, this matches what I've seen: about two years ago, kubernetes was at the height of hype, and enterprises were sniffing at it, but apps in prod and penetration were low. And even now, though penetration may be high (there's at least one use of kubernetes in an enterprises but maybe not, like, 100's it uses), there's still not wide use.
But, now, if Brandon is right, it's the right time to start putting it in place.
This further suggests a new time window: years to broad adoption. At first, most enterprises would take 3 to 6 months to understand then PoC a new app on a layer of infrastructure software- linger if they are considering, talking with, then getting through procurement with vendors.
If that goes well, they'll put more on there over the rest of the year. Then the next year, they'll put even more. So, let's say after two years of putting apps on kubernetes, for example, we'd say the enterprise is widely/significantly using it, based on number of apps deployed in production and a policy to be "kubernetes" first.
So, add that all up and you've got roughly 7 to 8 years for "full" adoption from the birth of a new infrastructure layer and, this, success of a layer of infrastructure software. (This excludes the initial lead time of creating the technology, e.g., the time Google spent to go from nothing to that release five years ago.)
The vendors (or open source communities - the distinction now is largely semantic) can accelerate this with good branding (Google did it, so it must be great!), hyping/evangelism (raising awareness), and getting/publicizing lots of enterprise references (if all these big banks are going it, it must be good).
Competitors will seek to slow this adoption with rival technology and FUD, or join in using the same tactics.
Your years may vary, or course. For example, Linux:
In August 2010, Jeffrey Hammond, principal analyst at Forrester Research, declared, "Linux has crossed the chasm to mainstream adoption," a statement attested by the large number of enterprises that had transitioned to Linux during the late-2000s recession. In a company survey completed in the third quarter of 2009, 48% of surveyed companies reported using an open-source operating system.
That's "mainstream" adoption. If you track "early success," you might track something like 20% marketshare instead of half the market - half the market would be "like, way successful). The iPhone and "mobile in the enterprise" probably have similar time lines. The iPhone was released in 2007, 12 years ago. I think we'd all agree that two to five years ago it'd "won," and mobile in the enterprise had long won: by 2016, mobile use was so high that even the most conservative and paranoid enterprise had to make it work, or try.
In general, then, a good rule of thumb for how long it takes a layer of infrastructure software to "win" is probably 7 to 10 years. Enterprises should consider new layer about two to three years after its birth, if the layer has built up enough hype and existing users.
All of that is a sliding scale, but it gives you a sense of how long all this takes: it's not months, even a year or two.
Original programming
Podcasts
This week we discuss Datadog, Ping, and VMware's kubernetes stuff - I abstain from the third. I also talked analyst relations with Richard and Rita. She's the best at what she does, as you'll hear.
Decks and talks
I gave a 90 minute version of my standard talk on Glasgow this week. Slides! I recorded myself as well and might could publish it.
Relevant to your interests
This year's DevOps Report is our - I took some notes while reading it. Revenue for low code companies getting into the $100m’s. Every enterprise software vendor wants to grow the amount each customer pays annually, make sure they increase use of the software, and get high level executives caring about the software as soon as possible. You always go back to MOM. Optimal app dev team size is small for most organizations: still around 8 people.
Coda
I'm trying, friends. This newsletter thing always dies. I've been at Pivotal five years and I keep thinking: that's five years I could have been doing a newsletter. I'm always inspired by Matt Levine's Money Stuff so endeavoring to steal his format, schedule, and verve.