Docker

Release / History

License: Apache 2.0
Started by Solomon Hykes at dotCloud
- Jeff Lindsay independent collaborator
Released as open source in March 2013
In the years 2013 – 2016 many organizations started to support Docker
- Cisco, Google, Huawei, IBM, Microsoft, and Red Hat
October 2015, the project had over 25,600 GitHub stars

LXC → libcontainer

Operating System Level Virtualization (no Hypervisor)
- based on Linux Containers (LXC)
- March 2014, with release version 0.9 Docker dropped LXC
Development of “libcontainer” started
- cross-system abstraction layer
- Docker execution environment
- builds up a new container specification
- wraps control-groups, namespaces and UnionFS
Similar implementations
- rkt, FreeBSD Jail, OpenVZ, system-nspawn

OS manages everything

a container is an isolation by defining limits through kernel features
- Namespaces
  - process ids, hostnames, user ids and network access isolation
- cgroups
  - Resource Management/-Limiting, CPU, memory, disk I/O, network
Docker introduces copy-on-write storage to manage images and containers
Docker aims to simplify and abstract this kernel level operations
a container feels like a VM but exists only due to an isolation mechanism provided by the OS

Docker as an abstraction layer

Software
- Docker daemon
  - listens to requests sent via Docker-Engine-API (REST)
- Docker client
  - uses Docker-Engine-API
Objects
- Images
- Containers
- Services
Registries
- Hub
- Sharing Images
- Public: Docker Hub, Docker Cloud
Platform independent Container Specification

Docker Images and Containers

Images
- built up from a series of layers
- each represents an instruction in a dockerfile or may be a manual configuration
- layers are read only, except the last one
Container
- a read write layer
- to be able to modify data which was read in an image layer, a copy-on-write process must be performed
- copy-on-write copies a file from an image layer to the container layer before the system can modify it
- all modifications may only be done on container layer
Flexible and Fast
- it is possible to start multiple containers of the same kind quickly
- versioning is easy to implement

Docker Volumes

Volumes
- define a volume which will be available in a container as folder
- managed by Docker daemon on host
- enables a sharing mechanism between containers
- flexible, as directly managed by Docker and thus platform independent
- remotely available
- may be started pre-populated in a container, to avoid copy-on-write
Bind Mounts
- directly mount a host folder into a container
- not managed by Docker, not platform independent
Tmpfs mounts
- temporary and only in memory
- useful for sensitive data
- no sharing, only on linux

Registry and Services

Registries enable sharing of images
- push images to a registry to start sharing
- pull images, to start from
- thousands of apps available, each in an single image
- different application versions, pre-configured in images
Public Registries
- Docker Hub, Docker Cloud
Services
- Scale containers across Docker daemons, swarm

Used to interact with Docker daemon

Docker-Toolbox
- works on Windows and Mac
- needs a virtualization driver to load a boot2docker iso
  - Virtualbox
  - Hyper-V
- provides Docker-machine command to remotely interact with the Docker host, which is running in a VM
DockerCLI
- provides commands to interact with the daemon, interacts mostly over Docker-Engine-API
Docker-Compose
- runs multi-container Docker applications
- configure a landscape using a YAML file
- define ports, networks, volumes, environment variables, start order, dependencies
Kubernetes
- Container Orchestration
- automated deployment, scaling, and management

Sources: