Ugh. This is what I get for posting late at night and trying to do so quickly. As a result, I don’t think I made my case very well. Let me address a few things by responding to SteppenWulf’s excellent critique of my post. Pardon me, but I’m going to quote his post a bit out of order for clarity.
The topic of the presentation you highlighted is container platforms - Cloud Foundry is not primarily a container platform but a vendor independent cloud tool set
What I was trying to avoid is having to ramble on forever trying to define CaaS and PaaS and what each thing means. Especially to what is probably a mostly non-technical audience. Because while CaaS is a fairly concrete term, the PaaS term gets thrown around quite a bit for a pretty wide category of things. Which I guess is why you prefer the term “independent cloud tool set”.
But, yes, I agree, I was 100% comparing apples and oranges.
- he talks about [CloudFoundry] in multiple places in the presentation starting around minute 4.
This is the thing that I really messed up about. The whole part around 4:00 was one of the things, perhaps the most important thing, I wanted to point out. Or, most specifically at 6:40. This is the core point I was trying to make. Let me actually post the slide here:
Traditional “PaaS” roles have been taken over by containers… The piece that is left for PaaS is the part that was always the most important part of PaaS in the first place, and that’s the opinionated developer experience. --Brendan Burns, Kubernetes cofounder.
Now the speaker makes it absolutely clear that this is not an unbiased opinion. But it is backed up from my own personal experience. The speaker has an interesting slide up around 14:55 where he shows the spectrum. He defines it as
IaaS → CaaS → PaaS → FaaS → SaaS
(FaaS isn’t really a term I’ve seen before, but I understand what he means.) But, in my personal experience, the Application platforms are getting squeezed. The left of the diagram gives you more control. The right of the diagram gives you more convenience. And I feel like the jump from CaaS to Pass is giving you less convenience than it used to, and the jump from FaaS/SaaS to PaaS doesn’t give you as much power as it used to.
3 years ago I was seeing 40% of people using hardware or IaaS directly, just hacking together a bunch of scripts, 30% using containers (mostly hacking together a lot of container management stuff), and 30% using PaaS. I knew a lot of people at Pivotal at this point, as well as at IBM Cloud Foundry.
But these days I’m seeing 20% using hardware/IaaS directly, 70% using containers, and 10% using PaaS. And this is the point that I was most interested in feedback on, because I am very aware it probably depends on where you work and in what groups.
Containers were very hot 2-3 years ago, but from my end, I see a lot fewer engagements using them over the last year or so. I think it is not clear yet whether containers or PAAS will win in the end. Containers require a lot more configuration work and upkeep, and that work doesn’t help the developers solve the business problem.
Which is why I find your response so interesting. I very well knew that you might have a different experience. But I’ve seen the exact reverse. 2-3 years ago I saw a lot of people really interested in containers, but very little hard core use because the container management features weren’t mature enough. Whereas, Pivotal had a really mature CloudFoundry platform that could be used for real stuff. So 2-3 years ago, I saw Pivotal with a real advantage. Now, it’s containers for anything of serious scale.
I think it is not clear yet whether containers or PAAS will win in the end. Containers require a lot more configuration work and upkeep, and that work doesn’t help the developers solve the business problem.
[Companies] always have lots of [developers] that want their own thing. Enterprise architects and the standards police try to corral them in to certain standard products and processes, but they never totally succeed.
These are two separate quotes from SteppenWulf’s post. But I think it illustrates my opinion. Yes, containers require more work to set up. But containers will work with basically everything. And that work can be automated. With the product I’m working with right now, I click a button and out pops a container. Click another button and that container is deployed in a clustered and elastic environment.
In order to use CloudFoundry, however, you do have to corral your developers a bit. Not everything works in CloudFoundry. CloudFoundry is “an opinionated developer experience”. In theory, the code I’m working with can deploy to CloudFoundry, but it’s a PITA because cloud foundry is “opinionated” about the network communication I need.
However that turns out, I don’t think the bull case on Pivotal is based on it having, in some way, the “perfect” or “best” technology. Technology is a moving target and is always changing. It’s built on the case that it has no significant competition in its very profitable market - a vendor independent cloud tool set in the enterprise space.
This is why I disclosed that I haven’t looked at Pivotal’s financials at all. Because I’m not sure what is priced in. Pivotal’s value prop is to go to those people in the 20% using scripts to deploy to IaaS and say to the IT managers “look at all the time that is being wasted”, “look at all of the tools I can provide you”, and “look at all of the errors I can prevent”. They then go to the developers with a similar message.
That’s not a bad value prop. But over time, I do think the growth is going to slow down. Other techniques (containers, non code-based platforms) are going to be more competitive. The market of people who are just starting in the cloud is getting smaller. If Pivotal is priced as a growth play, you are probably good. Pivotal has many good years ahead of it. If Pivotal is priced as a hyper growth play, about to hit a tipping point, I think that’s overly optimistic. I think Pivotal is past it’s hyper growth phase personally. (Although your anecdote is definitely evidence to the contrary.)
I think we’ll see a lot more specialization of containers before dust settles, and find that their best use if for custom edge cases. Even if 5 years from now we find out everyone prefers containers, I doubt there will be a single container that wins. Enterprise companies will still require vendor independent tool sets and processes.
Now for some interesting IT speculation. This has little to do with my case regarding Pivotal. Also, CaaS and PaaS are not my core competency. I write code. As much as possible, I try to let it be someone else’s decision and responsibility to deploy and manage that code. So take the following with a grain of salt. 100% possible there are winds blowing in this market I don’t know about.
But it really seems like the world is boiling down to Kubernetes as the base standard for container orchestration (and thus Docker for containers). The major cloud vendors will probably roll their own on top of that, and the enterprises will buy a stack of technology built on top of those two quasi-standards. The companies Docker and Kubernetes will do well, as the owners of those “standards” but will be squeezed by the players building value on top of them.
Just as Pivotal felt it important to have an open “CloudFoundry” and then leverage their early mover and innovator advantages to be the de facto best CloudFoundry implementation, I think we will have a container world that revolves mostly around Kubernetes. Yes, I know there are still some major exceptions to this, but Kubernetes seems to be that getting critical mass where anything not based on Kubernetes is going to have issues.
Also the speaker, Karl, does have an iron in the fire - he is working for a new container startup.
Yeah, I meant to point that out. Interesting to note that he worked for Pivotal previously as well.
I should also add my own disclaimer since I like to remain pseudo-anonymous on these boards. I do not work for any company with a direct stake in these markets. My current company does have relationships with several of these vendors (including Pivotal) because we want our product to run in these environments, but I mostly don’t care what our customers use. I’d be more specific, but I think it would provide too many clues to my identity.
I do think we probably have more companies using containers for deployment, which may explain why my perception of the market is so different from yours.
–CH