dockercon 2020–the end of containers

This is my 5th time at dockercon. It’s half empty, people here are not coming driven by excitement anymore but by the fact that they still have to maintain legacy container-based architectures that their employers invested in, sometime around 2015-2017.

The vibe and energy it used to radiate has dried up. The number of the booths have being halved compared to dockercon 2019.

Today the world moved into something completely different, valuing simplicity over the high cost of lock-in. What happened is that technology has shifted and we have skipped the containers madness all together to go directly into simple, auto-scalable micro-functions deployed in Serverless/FaaS platforms like AWS Lambda, Azure Functions or Google Cloud Functions. No more containers to manage, orchestrate, monitor and scale. Not much to learn or run, just micro-functions to deploy.

I first attended dockercon in 2016 in Seattle.

There was great momentum back then. Keynote in the general session was filled up with over 5,000 DevOps and IT leaders hungry to learn more and bring back some container-efficiency into their enterprises, following the vision of Solomon Hykes.

General Session

You could feel something was going to change the software industry forever and we would never go back. And indeed we never went back. Docker was leading the container revolution and had a growing community of new startups (our company was a sponsor in 2016), Fortune 500 and established partners jumping all over the places at the speed of light. It was well on its way of becoming a conference as big as AWS RE:Invent. Massive and unstoppable.

Datadog booth

Everyone wanted to learn more and be part of the movement. Expo area was filled up with over 50 booths from AWS, IBM, Cisco all the way down to new entrants like our own Kong (API Gateway) Datadog (Container Monitoring) or Crate.io (Container-based Scalable SQL).

Other tracks had very high quality contents from a large production usage scenario at Riot Games to Orchestrations and Microservice Architectures.

While a 9 year old kid was making a Docker demo!

It was the future.

What happened since then?

As mentioned before, over the last 5 years a new entrant and completely revolutionary concept took over and went mainstream: Serverless

Docker created the antidote of his own virus. It opened up a new way of “serving” the code: serverless and so the wind changed… It did not change overnight, it was slow.

Innovation happens on the edge first, at the margins of the culture and then expand backward from there. At first you tend to ignore it and classify as immature, then time passes by and more features show up, more companies join the small party and all of the sudden it goes mainstream and it becomes a giant rave you can’t stop. That is exactly what Docker did in 2013 to VMware and that is exactly what Serverless companies and platforms did in return to Docker.

I remember those early 2015-2016 days… At first most of the technology world didn’t want to adopt Serverless tools because they were lacking many essential features, there wasn’t any OSS or standard around them, no backward compatibility with code already written, no hyper-scale use cases and a strong fear of lock-in.

Unfortunately we all underestimated the new animal and the fact that it was just v 0.1.

Over the years Serverless tools added all the necessary features to build complex, secure, portable, enterprise-ready software. Banks (that I’m aware of) started to use AWS Lambda for side projects and then added more and more into their main core. Fast forward 2020 and Serverless/FaaS platforms are now running 50% of all the cloud-based software in the world.

We have skipped the container dream like we have skipped the SOA “SOAP- based” dream in the mid 2000. It became too complicated.

The world found a truly easy way to deploy, run and scale code, not only without managing servers or resources but also without managing (Kubernetes, DC/OS, Docker Compose) containers!

After thoughts:

I can see the docker team moving step by step into serverless tooling but I liked to paint a realistic future where we’ll be locked-in with the Lambda(s) of the world just because it was easier. This technology is easier to monetize as provider of Serverless/FaaS platforms (viable business, pay per use, charge the milliseconds, etc) and to adopt and run in production as a customer.

Docker has to keep an eye on the early serverless movement because it’s the real threat to their own revolution.