Short answer
Platform as a service (PaaS) is a cloud model in which the provider manages the infrastructure, operating system and runtime. You deploy and manage your application and data. PaaS has changed a lot since the first platforms appeared in the mid-2000s, and containers have been one of the biggest shifts. This article covers the main milestones, how containers changed the model, and how to build a small self-hosted PaaS with Docker.
For the three cloud service models side by side, see IaaS vs PaaS vs SaaS compared.
How PaaS developed
The first platforms (2006 to 2008)
Zimki, from the company Fotango, is often cited as one of the first PaaS products. It launched in 2006 and closed in 2007. Heroku followed in 2007, with support for Ruby first. Salesforce acquired Heroku in December 2010 for about $212 million.
Google App Engine was previewed in April 2008 with Python support. Java support followed in 2009. Windows Azure was announced in October 2008 and became generally available in February 2010. It started as a platform service and added infrastructure virtual machines later.
Standardizing the platform (2011)
Several platforms launched in 2011. AWS Elastic Beanstalk launched in January. Cloud Foundry came from VMware, and Red Hat OpenShift also arrived that year. The same year, Adam Wiggins of Heroku published the Twelve-Factor App methodology. It describes practices for building software that runs well on a platform, such as keeping configuration out of code.
Containers change the base (2013 onward)
Docker grew out of dotCloud, a PaaS company. Docker was open-sourced in March 2013, and dotCloud was renamed Docker, Inc. in October 2013. A container packages an application with its dependencies, so the same package can run in a test environment and in production.
Google announced Kubernetes in June 2014, and version 1.0 followed in July 2015. Kubernetes manages where containers run and keeps them running. On its own, it is not a PaaS. OpenShift version 3, released in 2015, was rebuilt on Docker and Kubernetes.
Cloud Native Buildpacks, a project under the Cloud Native Computing Foundation, appeared in 2018. It builds container images from source code without needing a Dockerfile.
Pricing has also changed. Heroku ended its free plans on November 28, 2022, which is a reminder to check current pricing before you commit to a platform.
Did containers replace PaaS?
Not really. A common view is that containers took some PaaS ideas, such as packaging and deployment, and made them portable enough to run on your own terms. Many managed platforms now run applications in containers.
What a PaaS gives you, and what it does not
A PaaS can speed up deployment, handle scaling and take over routine tasks such as operating system patching. Pricing is often pay as you go.
A PaaS also has trade-offs. You have less control over the underlying servers, and some platform limits are set by the provider. Moving between platforms can be hard, because each one has its own services and conventions.
Why build your own PaaS?
A self-hosted PaaS is a reasonable choice for a few reasons: you want control over where your apps run, you want to keep costs predictable on a single server, or you want to learn how platforms work.
The trade-off is responsibility. With a self-built platform, you still handle operating system patching, backups, monitoring and security. A managed PaaS would do much of that for you.
The building blocks
A small self-hosted PaaS usually combines these parts:
- A virtual private server (VPS) with a public IP address
- Docker to build and run each app as a container
- A reverse proxy such as Traefik, Caddy or nginx, which routes domain names to the right container
- TLS certificates from Let’s Encrypt, renewed automatically. Let’s Encrypt certificates last 90 days.
- A deployment trigger, such as a git push hook or a CI pipeline that builds the image
- A subdomain for each app, so every app has its own address
Building it step by step
This outline assumes a Linux server and basic command-line experience.
-
Prepare the server. Create a non-root user, set up SSH keys, install updates, and open only the ports you need, usually SSH, 80 and 443.
-
Install Docker. Follow Docker’s official installation guide for your distribution.
-
Set up the reverse proxy. Run a proxy that can request and renew TLS certificates automatically, such as Caddy or Traefik. Point your domain names at the server.
-
Build the app image. Add a Dockerfile to your project and build the image:
docker build -t myapp . -
Run the container. Bind the port to the local interface, so only the reverse proxy can reach the app:
docker run -d --name myapp -p 127.0.0.1:3000:3000 myapp -
Automate deployments. Trigger a build on each git push, or have CI build the image and restart the container.
-
Add monitoring and backups. Check logs, alert on disk and memory use, and test that you can restore your data.
Ready-made self-hosted options
If you would rather not build the glue yourself, several open-source projects provide a PaaS-style workflow on your own server:
- Dokku is a small, Heroku-style platform. You deploy with a command such as
git push dokku main, and it builds your app with buildpacks or a Dockerfile. - CapRover, Coolify and Dokploy are self-hosted platforms with web dashboards.
Check each project’s current documentation for requirements and features before you choose one.
Security checks before you go live
- Do not mount
/var/run/docker.sockinto a container. Access to the Docker socket is root-equivalent on the host. - Do not expose the Docker daemon on TCP port 2375. It has no authentication or TLS by default.
- Open only the ports you need, and keep the host operating system, Docker and your images updated.
When a managed PaaS is the better choice
A DIY platform is a good learning project and can suit small, low-stakes workloads. For production systems where uptime and patching matter, a managed PaaS often means less work for your team. Managed platforms come with less control and possible vendor lock-in, so weigh those trade-offs for your situation.
For a related managed option, see what backend as a service is, which is usually classed as a subset of PaaS. For managed databases, see database as a service.
Frequently asked questions
What is the meaning of PaaS in cloud computing?
PaaS stands for platform as a service. The provider runs the infrastructure and runtime, and you deploy and manage your application and data. Examples include Heroku, Google App Engine and AWS Elastic Beanstalk.
Is Gmail a PaaS or SaaS?
Gmail is SaaS. It is a finished application that people use through a browser or app, and Google runs the whole stack. PaaS is for platforms where developers build and run their own applications. To classify other services, see IaaS or PaaS? Azure Storage, App Engine and Elastic Beanstalk.