Test the ABM in docker/swarm

[chintanp]

• Use the docker version GAMA
• Deploy one agent / family per container.
• Provision on AWS such that each container has access to one CPU.
• Provision the DB for a high ingestion rate if it proves to be a bottleneck.
• Save the instantaneous state to the database, and read the state from database continuously for EVSE agents, or use docker exec to communicate with other containers.

[chintanp]

• The EVs are the one that take most of the time. Several decisions to be made before a step is taken.
• Distribute EVs across containers / GAMA processes.
• The only points of conflict are:
  • State queue_for_charging
  • *State charging (nearest_evse.plugs_in_use)
  • *State waiting
• At these points of conflict, refer to the collected value from the DB. Consider using a database trigger to get value from the database. And this is a point of contention. Each container has a different agent, with different roads, capacity, etc. So they are likely not going to be time-synchronized. i.e. at 10.43 AM could happen in 2secs on container one and 2 minutes in container 2.

[chintanp]

• Maybe artificially synchronize the processes. Force the going to next step on a database signal, which can wait for a particular insert from all agents.
  • agents are ok to die in their lifetime. The signal will ensure to count dead+alive agent inserts count.