bragov4ik/p2p-gossip

Test task

P2P gossip application

Description

This is an application for creating and connecting to a simple peer to peer network. Nodes in the network periodically ping all other members and receive (and display) others' pings.

How to use

To launch the node, you need to run it with appropriate command line arguments. List of available args you can get with the use of --help/-h or from the list below.

For example, to start the first node in the network, you can use

p2p-gossip --period 5 --port 8080 It starts a node that will ping others each 5 seconds and listen for new connections on TCP port 8080. Its keys will be randomly generated and stored in program memory.

p2p-gossip --period 3 --port 8081 --connect 127.0.0.1:8080 This will start a node that will ping others each 3 seconds, accept connections on port 8081, and will try to join the network through the node at 127.0.0.1:8080 at the startup.

Logging levels are controlled by env variable RUST_LOG. You might want to set it to RUST_LOG=off if you don't want to get read any logs (for demo for example).

Arguments description

• --period (optional) time between ping messages sent to other peers. Default is 5.
• --port (optional) port at which the node will listen to new incoming connections. Default is 8080.
• --connect (optional) address through which to join the network. If nothing provided, just wait for incoming connections.
• --key (optional) path to file with private key. If not provided, generates a new one.
• --cert (optional) path to file with certificate (public key). If not provided, generates a new one.

Network behavior and summary

This is a fully-connected network, which means all nodes directly communicate to each other. This allows to simplify ping distribution process, since each peer is responsible only for itself.

Joining

Workload on joining the network lies on the new node. The process is the following:

1. Communication with the first peer
   1. The node connects to some peer in the network
   2. They exchange their information and add each other to known peers list
   3. The node asks for list of peers and gets a result
2. Communication with the remaining peers
   1. The node connects to all other nodes from the list
   2. They exchange info, add themselves to their lists
   3. If any other nodes found, repeat from 2.1

Authentication and (some) security

Nodes have identity that consist of their hashed public keys. When a node joins the network, peers simply add its identity to their known peers list. On the following reconnections to the network, the peers receive the same identity and match it to given public key, authenticating the node this way (in fact, the node also verifies other peers the same way).

The traffic is encrypted using TLS with certificate verification procedure described above, i.e. known peers' public keys are compared with their identities and new peers are simply accepted and rememebered.

Status check

Network uses heartbeat messages for checking status of other nodes. If a node doesn't receive a heartbeat from another for some time, it considers the peer dead.

Connection loss

If a node losses connection to the peer (either by closing TCP connection or absence of heartbeat) it starts reconnection procedure. On reconnection, the peers authenticate each other and exchange lists of known peers. It is done for the case of network splitting, when new nodes are added to both sides.

Possible further improvements

• Store known peers in persistent storage to reconnect after shutting application down
• Maybe better authentication
• Better logging