D'Autilio Joel - Rossi Pablo
This project was developed as the final project for the Telecomunications and Distributed Systems Course at Universidad Nacional de Río Cuarto, Argentina.
The Chord algorithm is designed to provide efficient and scalable key-value lookups in decentralized environments. It utilizes a distributed hash table (DHT) to distribute data uniformly across nodes, ensuring balance and scalability. Chord works well for dynamic networks where nodes may frequently join or leave.
In this project, we apply the Chord alogorithm to implement a distributed users database. The system handles a set of Users, and each node contains a subset of it. A node can fetch a user's information (from it's own table or from one if its peers), or it can insert/update an user.
This implementation also supports fault tolerance, ensuring the system remains operational and no information is lost even if a number of nodes fail.
To install the project, it's recommended to first create a virtual environment for python. For example, one can be created and activated using venv with the following commands:
python -m venv .env
source .env/bin/activate
After that, the dependencies can be installed with the following:
pip install -r requirements.txt
Once the dependencies are installed, the peer.app script can be used to launch an HTTP server along with a GRPC server. The arguments can be found with this command;
python -m peer.app -h
This command runs the HTTP server with port 5000 and the GRPC server with port 50051, both in localhost.
python -m peer.app -p 50051 -P 5000
To join to an existing ring, to a chord node also in localhost, the following command can be used.
python -m peer.app -p 50052 -P 5002 -j 50051
If the target chord node to join to is in a different machine with ip <join-ip>, use this command.
python -m peer.app -a <my-ip> -p <chord-port> -P <http-port> -J <join-ip> -j <join-port>
Once the ring is stablished, the setusers script can be used to add multiple users to the system. For example, to add 10 users through a node in this machine with port 5000, use the following command:
setusers -n 10 -p 5000
The Users table in a local node are displayed in output/<port>/users.txt.
In the same directory, the log.txt file displays the chord node's log, and finger.txt contains internal information (ID, predecessor, finger table).
The chord server implementation can be found under peer/.
Inside it, chord/ and users/ contain the code for the GRPC chord and users servicers.
The former focuses on the infrastructure for the server, handling node joins, leaves, finding and stabilization processes. The later acts as a database manager, handling the addition, removal and transference of users.
The proto/ directory contains the protobuf definitions for the remote call procedures (RCP) used by both servicers. The code generated by GRPC is placed under stubs/.
The peer/server.py file contains the Server class, implementing a GRPC server which uses both servicers previously mentioned.
The peer/app.py script launches a GRPC server along with an HTTP server, exposing two endpoints (GET and POST) to fetch and insert users.
.
│ constants.py
├───peer
│ │ address.py
│ │ app.py
│ │ server.py
│ ├───chord
│ │ chord_servicer.py
│ │ node.py
│ │ remote.py
│ │ utils.py
│ └───users
│ remote.py
│ users_servicer.py
│ utils.py
├───proto
│ chord.proto
│ users.proto
└───stubs
...
The following diagram visualises the implementation of a peer node.
