####Scheduler Swarm comes with a builtin scheduler. It currently provides basic functionalities, such as scheduling containers based on constraints or affinity (co-scheduling of containers), persistent storage and multiple scheduling strategies like binpacking or random.
We plan to add more features to the builtin scheduler such as rebalancing (in case of host failure) and global scheduling (schedule containers on every node)
- Virtual Container ID
- Rebalancing
- Global scheduling
####Leader Election (Distributed State) Regarding Swarm Multi-tenancy, we are working on shared states between multiple "soon-to-be-master" and master election.
- Shared state
- Master election
####API Matching Swarm currently supports around 75% of the Docker API as you can see here
Our goal is to support 100% of the API, so all the Docker CLI commands would work against Swarm
- Bring Swarm API on par with Docker API as much as possible
- Use labels instead of env variable for constraints and affinity
- Open a few PRs on docker/docker to display labels in
docker psanddocker images
####Networking Scheduling and container placing is only part of the problem. In order for Swarm to be a good solution for distributed environments, we need to support multi-host networking.
- Integrate Swarm with Docker Networking.
####Extensibility The builtin scheduler allows you to manage container on approximately hundreds of machines within a cluster. In the future, we want to allow third party tools to be plugged into Docker Swarm to allow management of much bigger clusters.
The first integration will be Mesos but other tools will arrive after, like Kubernetes.
- Cluster Drivers #393
- Mesos implementation
An Open-Source Planning Process is used to define the Roadmap. Project Pages define the goals for each Milestone and identify current progress.