Distributed System with Ring Topology

A distributed system implementation featuring a ring topology with resource management and deadlock detection capabilities. The system uses both REST API and RPC for inter-node communication.

Features

• Ring topology with self-healing capabilities
• Distributed resource management
• Chandy-Misra-Haas deadlock detection algorithm
• REST API for external control
• RPC for inter-node communication

Architecture

Node Structure

Each node maintains connections to:

• Next node
• Next-next node (for topology repair)
• Previous node

Communication

• REST API on port N+1 for external control
• RPC on port N for inter-node communication
• Binary protocol using tarpc for RPC

Building and Running

Prerequisites

• Rust toolchain
• Cargo

Building

cargo build --release

Running a Node

./target/release/distributed_system <ip> <port> <resource_name>

Example:

./target/release/distributed_system 127.0.0.1 2010 A
./target/release/distributed_system 127.0.0.1 2020 B
./target/release/distributed_system 127.0.0.1 2030 C

Control Script

A bash script (control.sh) is provided to interact with the system:

chmod +x control.sh
./control.sh

Available commands:

• g <idx> - Get node health
• s <idx> - Get node status
• j <from_idx> <to_idx> - Join nodes
• l <idx> - Node leaves
• k <idx> - Kill node
• r <idx> - Revive node
• acq <idx> <resource> - Acquire resource
• rel <idx> <resource> - Release resource
• det <idx> - Start deadlock detection
• wait <idx> <target_idx> - Wait for message
• active <idx> - Set node active
• passive <idx> - Set passive
• delay <idx> <ms> - Set delay

Example Usage

Creating a Ring Topology

# Start 3 nodes in separate terminals
./target/release/distributed_system 127.0.0.1 2010 A
./target/release/distributed_system 127.0.0.1 2020 B
./target/release/distributed_system 127.0.0.1 2030 C

# In control script:
j 1 0  # Join node 0 to 1
j 2 1  # Join node 1 to 2

Implementation Details

Resource Management

• Resources are uniquely identified by strings
• Each resource has one owner
• Resource requests are queued when busy
• Resource state includes current user and request queue

Deadlock Detection

• Implements Chandy-Misra-Haas algorithm
• Uses probe messages for cycle detection
• Supports active/passive state transitions
• Detection can be initiated from any node

Fault Tolerance

• Self-healing ring topology
• Node failure detection
• Topology repair mechanism
• Resource state recovery

License

This project is licensed under the MIT License - see the LICENSE.md file for details.