Mokka Consensus Algorithm implementation in Node.js.
Concept description (PDF)
Live Demo (in browser)
consensus features
- resistant to network splits
- non-anonymous voting
- voting validation with musig
implementation features
- Custom transport layer support: Mokka separates interface implementation and consensus
- Fully customizable: you can create your own state machine around Mokka consensus (check out demos for more info)
- Can run in CFT and BFT modes
$ npm install mokka --save
$ npm run build
Returns a new Mokka instance. As Mokka is agnostic to protocol implementation,
you have to create your own.
Please check the Custom transport layer
section.
Arguments:
address
(string): an address in custom format. The only rule is that address should include the public key in the end (example:"tcp://127.0.0.1:2003/03fec1b3d32dbb0641877f65b4e77ba8466f37ab948c0b4780e4ed191be411d694"
)crashModel
("CFT" | "BFT"
): crash model, which should run the consensus. The difference is in quorum - CFT requiresf + 1
nodes for quorum, while BFT2f + 1
heartbeat
(integer): leader heartbeat timeoutelectionTimeout
(integer): candidate election timeout (i.e. vote round)customVoteRule
(func): additional voting rulereqMiddleware
(func): request middleware (will be triggered on every new packet received)resMiddleware
(func): response middleware (will be triggered on every new packet sent)proofExpiration
(integer): when the leader's proof token should expire.logger
(ILoggerInterface): logger instance. If omitted, then console.log will be usedprivateKey
: the 64 length private key. Please take a look at example key pair generator
Add new peer node by uri
Start consensus. Should be called after all nodes has been added.
Create new packet, where type
is packet type, and data
some custom data
Decode packet from buffer
Send message to peer
A Mokka instance emits the following events (available at /components/shared/EventTypes.ts
):
join
: once we add new peerleave
: once we remove peerheartbeat_timeout
: once we can't receive the heartbeat from leader in certain time (specified in config)state
: once the state of node changed (i.e. leader, candidate, follower)
Mokka is a log-less consensus algorithm and doesn't provide any RSM (i.e. replicated log). You have to implement your own. However, there is a good example of RSM implementation, which is similar to RAFT.
In order to communicate between nodes, you have to implement the interface by yourself. As an example you can take a look at TCP implementation: src/implementation/TCP
.
In order to write your own implementation you have to implement 2 methods:
-
The
async initialize()
function, which fires on Mokka start. This method is useful, when you want to open the connection, for instance, tcp one, or connect to certain message broker like rabbitMQ. -
The
async write(address: string, packet: Buffer)
function, which fires each time Mokka wants to broadcast message to other peer (address param).
Also, keep in mind, that Mokka doesn't handle the disconnected / dead peers, which means that Mokka will try to make requests to all presented members in cluster, even if they are not available. So, you need to handle it on your own.
Node.js |
---|
running cluster |
running private blockchain |
Node.js |
---|
TCP |
ZMQ |
However, you still can implement your own protocol.
Copyright (c) 2018-2021 Egor Zuev