The narrative seems quite clear to me. They released the tooling and the services to become the defacto solution, and then Swarm was supposed to be the cashcow that turned that into cashflow. And then k8s happened.
They've raised a tonne of capital, and it probably looked sane at the time. And now they're grasping at straws trying to figure out how else they can turn this into revenue.
A lot of the recent narrative has been worded like pivoting into a glorified webhost was their evil plan all along. It's not, it's an act of desperation.
I feel like all that VC money was actually their undoing. Instead of working with the community, Docker spent bucketloads of money acquiring a lot of community projects and startups in an obvious attempt to become an end-to-end container solution company. But, they didn't have a plan and ended up killing most of those acquisitions. To me, it felt like Docker was using all the money they got to squash the community instead of working with it.
Every year at DockerCon, there would be flashy announcements that went nowhere. As a developer, those years from 2013 to 2017 were both super exciting and super frustrating. Everything started falling apart when Docker (the project) got split into Moby for open source and the rest went commercial. Docker started to sell Docker Swarm (the original), only to kill it a year later with a new Docker Swarm (what we have today). Then, Kubernetes started growing traction, leapfrogging both Docker Swarms, Mesos, and others in adoption. They never had a cohesive commercial plan. Just lots of empty promises and burned bridges.
When I think of Docker (the company), I feel bitter about all the projects they killed in their attempt to own the market. I love using Docker (the software), but the company's just one big disappointment.
>I feel like all that VC money was actually their undoing.
I don't think you have Docker without VC money. I think at best you have some LXC-lite project that is intertwined with GCP/AWS/Azure but unfortunately, I don't think you get something as quite polished as Docker in the same timeframe without VCs pouring millions to hire people to work on it.
Where I am sympathetic to Docker is Docker wouldn't have worked as your standard Open Source project; it took a ton of paid engineering hours (and you can argue X% was wasted on projects that went nowhere, but that's a given in any org), to get the software and infrastructure right, and if they had tried to charge developers they would have gotten nowhere. Even now, where people are keenly aware of the value of Docker, trying to monetize it is met with tears and angry blog posts.
I think Swarm was the plan (as the money has always been in providing infrastructure), Google just had more developer sway (which also killed Mesos).
The way I see it are there probably a ton of services you could build with the right team and right amount of people that could make an even greater impact than Docker on productivity and they would never be built because it would be way too difficult to monetize and the people with the talent to build it are going to get paid more working at FAANG to do any sort of OSS approach.
>there probably a ton of services you could build with the right team and right amount of people that could make an even greater impact than Docker on productivity and they would never be built because it would be way too difficult to monetize
There's a big difference between creating value and capturing value. Probably thr biggest single issue with capitalism in general.
In all fairness, Docker never worked with the community at all. At times it looked like they almost got their own community going, but it was never a priority to them.
They were always in a race to reinvent everything the Docker way. Which is, sort of, what you need to do if you want to become an enterprise software company, if that's where you think the money is. It's pretty much exactly in the footsteps of VMware. But to make that work, they would have had to work much closer to Windows, which probably was even harder than the Linux community.
Docker couldn't have existed without VC money. Free hosting was what made them hugely popular. Docker was a VC productization of containers.
Speaking personally, the fate of Docker, Inc. was clear to me when they took their $40M Series C round in 2014. I had met with Solomon in April 2014 (after their $15M Series B) and tried to tell him what I had learned at Joyent: that raising a ton of money without having a concrete and repeatable business would almost inevitably lead to poor decision making.
I could see that I was being too abstract, so I distilled it -- and I more or less begged him to not take any more money. (Sadly, Silicon Valley's superlative, must-watch "Sand Hill Shuffle" episode would not air until 2015, or I would have pointed him to it.) When they took the $40M round -- which was an absolutely outrageous amount of capital to take into a company that didn't even have a conception of what they would possibly sell -- the future was clear to me.
I wasn't at all surprised when the SVP washouts from the likes of VMware and IBM landed at Docker -- though still somehow disappointed when they behaved predictably, accelerating the demise of the company. May Docker, Inc. at least become a business school case study to warn future generations of an avoidable fate!
It always seems like VCs. They did great and started growing a lot. More VCs jumped into the next possible “unicorn” and they had even more money. They had to do stuff with it.
But they weren’t paying back fast enough, and K8s was making waves. Better make money fast while you can. Time for the squeeze play so the VCs can win.
If they had been allowed to grow at a more natural rate maybe it would all be fine. If they were allowed to be happy with a 40% share of a big future market, things could be great.
Yeah, that's the elephant in the room that the article claims to acknowledge, but misses by an order of magnitude.
"Docker is also a company, and companies are expected to produce revenue." That reads like wages, office rent and perhaps a boat for the founders. But they can't be that humble anymore, not after burning through a quarter billion in VC money. (even if those numbers seems surprisingly harmless compared to some more consumer oriented VC bonfires)
Now it's their job to make those investments winning bets or ruin the company trying. All that free image CDN convenience much of the docker revolution was built on? Too good to be true, effectively a loan from VC betting on getting people hooked.
We are still using Docker Swarm in production. It seems to be working fine so I always wondered why it never took off. But I am not a Devop. Can sb please give some insight on why kubernetes took off instead and why Dock Inc. failed with its cloud product?
k8s is basically a standard interface for cloud providers, where you can define load balancers, persistent volumes, certificates, etc. in addition to jobs and services. It has a well thought out declarative language that maps well to how ops runs things at scale.
There are quite a few optional tools built on the k8s model, like service meshes or tools to orchestrate a postgres cluster.
Swarm is a small piece of that, the scheduling system, and you have to run it yourself. It's really not comparable.
"Can sb please give some insight on why kubernetes took off instead and why Dock Inc. failed with its cloud product?"
IMHO, Google did something smart/evil here. Google did not use the docker-compose style yaml files for k8s configurations. This turned into a big deal. If you have a docker-compose file and want to run it in k8s - you cannot. So you need to make a decision: go with docker-compose yamls which has no place to be hosted except swarm OR go with k8s yaml style and be able to run it on aws,gcp, etc.
I still use Docker Swarm in production, but the main issue is the lack of support from cloud provider and tooling. For instance, there is no good solution like ArgoCD for continuous deployment and updating your stacks. In addition, if you want to be able to automatically add a new node to the cluster, you basically have to write some infrastructure code by yourself. This makes infrastructure as code / configuration as code a pain, and these things are required to be compliant with stuff like SOCII
Good question, I'm also interested in the answer. Probably has something to do with K8s being backed by Google, which has battle tested it in the form of their internal Borg infrastructure.
My understanding is that k8s was definitely never used internally at google (at least at the beginning). They may have learned some lessons from Borg, but k8s is very very much not Borg nor a Borg component. I remember lack of battle testing and architectural differences being a key criticism when k8s was released.
Swarm is fine but k8s is a lot more advanced. I've used both and now that I understand k8s, I would choose it over swarm any day. Microk8s is neat if you want a lightweight cluster.
They've raised a tonne of capital, and it probably looked sane at the time. And now they're grasping at straws trying to figure out how else they can turn this into revenue.
A lot of the recent narrative has been worded like pivoting into a glorified webhost was their evil plan all along. It's not, it's an act of desperation.