(Image from aforementioned EIP)

Transaction module and block processing module are not implemented here since the aim of the project is to visualize the consensus.

Each node contains two objects

1. p2p - which is backed by libp2p mdns and swarm
2. state - which is implemented for messages handling

since it's using mdns peer finding behavior, so it can be only used in LAN.

The swarm is responsible for peers finding and messages passing.

• When a peer subscribes the message channel, it will call state.on_new_peer(peer_id) to check if it's able to start(the count of nodes >= 2f+1).
• when a message is passed through the channel, it will call state.on_message(msg, peer_id) to process the message and change phase if applicable.

Both on_new_peer and on_message generate responses, when the responses are marked with BROADCAST, the messages in it will be broadcast to the whole network. The two functions fullfill the needs of normal phase transition.

For RoundChange, we need to set a ticker to check if the node needs to change round, the ticker will call state.check_timeout() every 200ms.

For the state of one node, there is a phase attribute, every handler on_message, on_new_peer, check_timeout, ... will check current phase to determine if it's able to go on. If it's not in right phase, the node will simply ignore the message.

You need to configure the Mysql credentials at main.rs L71, then create a table in pbft database called edges as

The project is backed by rust, you need to install rust first and have cargo installed. After installation, type

./scripts/start.sh <f>

for simple visualization, just set f to be 1.

If want to test RoundChange functionality, set f to be 2 and close the window of Node 0 (actually random one that with id<2f+1), you will find all the nodes will wait 5s when primary fails, and continue execution after RoundChange if the new node is not faulty. If the new node is faulty, they will continue RoundChange till find a good one.

f=1

f=2