fawind / mongochain Goto Github PK

• Only one node should be the leader node.
• Initially, we should set this through our docker config (e.g. the seed node also runs the leader)

Setup simple mocked consensus service for testing communication and general end2end functionality. This gets replaced by the PBFT consensus service.

Some possible "forced" consensus options:

• Lowest checksum
• One hard-coded node broadcasts result
• [...]

Incoming transaction are put into a backlog. The backlog is then used by the consensus service to get pending transactions.

Don't use the default Java serialization for Akka messages. This results in bad performance and lots of warnings.

To replace #8.

Consensus service using PBFT protocol. This service is used for the inter-committee consensus.

Tasks:

• Research if there are existing solutions we can use.
• Implement consensus round.
  • Initially, sketch out outline of the flow.

• Configure env vars when running the service locally without docker.
• This should start a single node with a fault threshold of 0 for local testing.

• Pull logs from all cluster (scp, digital ocean storage?)
• Webservice for counting community ids and primaries
  • Start id at 0, count up using modulo, first id is primary
• Add autostart script for cluster nodes
• Modify test runner to read ips from input file

Setup node deployments using docker. This allows us to easily run multi node experiments.

Goal: Generic API for sending and receiving messages using PubSub. Messages are serialized objects.

Tasks:

• Method to broadcast message to topic.
• Method to subscribe to topic.
  • Should accept a callback function as parameter which gets invoked on incoming messages.
• Object serialization and deserialization for messages.

Parse performance metrics from logs

Things to log:

• Transaction life-cycle (start, stages, consensus)
• Queues and Message Box sizes

TODO:

• Create log events for easy parsing
• Add pipeline for pulling and unifying logs from docker containers
• Add scripts for calculating metrics (e.g. average consensus duration, ...)

1. Consensus round within committee.
2. Consensus round within final committee.

Placeholder issue for hosting image resources.

Dynamically assign nodes to committees based on proof-of-work.