p2p

P2P and Distributed Computing Guides

Comprehensive guides for peer-to-peer networking, distributed workflows, and libp2p integration.

Quick Links

• P2P Setup Guide - Main P2P setup guide
• P2P & MCP Architecture - Complete P2P and MCP documentation
• P2P Workflow Scheduler - Workflow scheduling guide
• P2P Workflow Quick Start - Quick start guide

Specialized Guides (in this directory)

P2P Cache

• P2P Cache Deadlock Fix - Troubleshooting deadlocks
• P2P Cache Encryption - Secure P2P communications

P2P Workflow Management

• P2P Workflow Discovery - Discover and join P2P workflows
• P2P Workflow Scheduler - Advanced scheduling features

Autoscaling

• P2P Autoscaler Quick Reference - Quick reference for autoscaling

libp2p Integration

• libp2p Universal Connectivity - Universal connectivity setup
• MCP P2P Setup Guide - Integrate MCP with P2P

Core Concepts

Peer-to-Peer Networking

IPFS Accelerate uses libp2p for peer-to-peer networking, enabling:

• Distributed Inference: Share compute across network peers
• Content Addressing: Cryptographically secure model distribution
• Fault Tolerance: Automatic failover and recovery
• Load Balancing: Distribute work across available peers

Merkle Clock Consensus

The P2P workflow scheduler uses Merkle clocks for distributed consensus:

from ipfs_accelerate_py.p2p_workflow_scheduler import MerkleClock

# Create clock for this node
clock = MerkleClock(node_id="node-123")

# Synchronize with other nodes
clock.update(other_node_clock)

# Get consensus hash
consensus_hash = clock.get_hash()

Fibonacci Heap Scheduling

Efficient priority-based task scheduling:

from ipfs_accelerate_py.p2p_workflow_scheduler import P2PWorkflowScheduler

scheduler = P2PWorkflowScheduler(node_id="worker-01")
await scheduler.start()

# Submit high-priority task
await scheduler.submit_workflow({
    "name": "urgent-task",
    "priority": 1,  # Lower = higher priority
    "tasks": [...]
})

Quick Start

Basic P2P Setup

# Install P2P dependencies
pip install ipfs-accelerate-py[p2p]

# Start P2P node
ipfs-accelerate p2p start --node-id my-node

# Join network
ipfs-accelerate p2p join --bootstrap /ip4/...

Distributed Workflow

from ipfs_accelerate_py.p2p_workflow_scheduler import (
    P2PWorkflowScheduler,
    WorkflowTag
)

async def main():
    # Create scheduler
    scheduler = P2PWorkflowScheduler(node_id="worker-01")
    await scheduler.start()
    
    # Submit distributed workflow
    workflow_id = await scheduler.submit_workflow({
        "name": "batch-inference",
        "tag": WorkflowTag.P2P_ELIGIBLE,
        "tasks": [
            {"model": "bert-base", "input": "text1"},
            {"model": "bert-base", "input": "text2"}
        ]
    })
    
    # Monitor progress
    status = await scheduler.get_workflow_status(workflow_id)
    print(f"Progress: {status['completed']}/{status['total']}")

Network Architecture

┌─────────────────────────────────────────────┐
│         P2P Network Topology                │
├─────────────────────────────────────────────┤
│                                             │
│  ┌──────────┐      ┌──────────┐            │
│  │  Node A  │◄────►│  Node B  │            │
│  │ (Worker) │      │ (Worker) │            │
│  └──────────┘      └──────────┘            │
│       ▲                 ▲                   │
│       │                 │                   │
│       ▼                 ▼                   │
│  ┌──────────┐      ┌──────────┐            │
│  │  Node C  │◄────►│  Node D  │            │
│  │(Scheduler)│      │(Bootstrap)│           │
│  └──────────┘      └──────────┘            │
│                                             │
└─────────────────────────────────────────────┘

See Also

• Main Documentation
• P2P & MCP Architecture
• GitHub Guides
• Deployment Guides

---

Last Updated: January 2026