Paxos Banking System with SmallBank Benchmark

A distributed banking application implementing the Paxos consensus protocol for state machine replication, with comprehensive performance evaluation using the SmallBank benchmark.

Overview

This project implements a fault-tolerant distributed banking system using the Paxos consensus algorithm. The system ensures that all banking transactions are consistently maintained across multiple nodes (replicas) through consensus-based state machine replication.

Features

• Paxos Consensus Protocol: Implements leader election, normal operations, and node failure handling
• Distributed Banking: Supports 10 clients (A-J) with transfer transactions
• Fault Tolerance: Handles leader failures, node isolation, and recovery
• Checkpointing (Bonus 1): Periodic state saving for efficient recovery and log optimization
• SmallBank Benchmark (Bonus 2): Comprehensive performance evaluation framework

System Architecture

• 5 Paxos Nodes: Distributed across ports 8001-8005
• 10 Banking Clients: Clients A through J with initial balance of 10 units each
• Leader-based Consensus: Single leader coordinates all transactions
• Majority-based Decisions: Requires 3 out of 5 nodes for consensus

Key Components

Core Paxos Implementation (main.py)

• Leader election with prepare/promise messages
• Transaction processing with accept/accepted/commit phases
• Node failure detection and recovery
• State synchronization and catch-up mechanisms
• Checkpointing system for efficient state recovery

SmallBank Benchmark

• smallbank_benchmark.py: Generates realistic banking workloads
• benchmark_evaluator.py: Performance metrics collection
• run_benchmark.py: Comprehensive benchmark suite
• 6 Transaction Types: Amalgamate, Balance, DepositChecking, SendPayment, TransactSavings, WriteCheck

Performance Results

The system achieves excellent performance with the SmallBank benchmark:

• Maximum Throughput: 330.50 transactions/second
• Perfect Reliability: 100% success rate across all tests
• Linear Scaling: Throughput scales with workload size
• Consistent Execution: ~2.6 seconds regardless of workload

Usage

Run Basic Tests

python3 main.py tests/input4.csv -d

Generate SmallBank Workload

python3 smallbank_benchmark.py

Run Performance Benchmark

python3 run_benchmark.py --test throughput

Run Comprehensive Benchmark

python3 run_benchmark.py --test all

Test Scenarios

The system handles various scenarios including:

• Normal transaction processing
• Leader failures and elections
• Node isolation and recovery
• Concurrent leader elections
• Insufficient nodes for consensus
• State synchronization after failures

Files

• main.py - Core Paxos implementation
• smallbank_benchmark.py - SmallBank benchmark implementation
• benchmark_evaluator.py - Performance evaluation framework
• run_benchmark.py - Benchmark runner
• tests/input4.csv - Test scenarios

Bonus Implementations

Bonus 1: Checkpointing

• Periodic state saving after every 100 committed requests
• Efficient recovery using checkpoints instead of processing all previous requests
• Log optimization with checkpoint-based catch-up mechanisms
• State synchronization for failed nodes using checkpoint data

Bonus 2: SmallBank Benchmark

• Comprehensive performance evaluation using industry-standard benchmark
• Realistic banking workloads with 6 transaction types and skewed access patterns
• Performance metrics including throughput, latency, and consistency
• Scalability testing with varying workload sizes and failure scenarios

Academic Context

This project implements the Paxos consensus protocol as described in the course materials, with additional SmallBank benchmark evaluation following the paper "Serializable isolation for snapshot databases" by Michael J Cahill, Uwe Röhm, and Alan D Fekete. At the end I have used AI assistance for creating this readme and understanding the benchmark behavior and context.