coosld/ceremonyclient
1
0
Fork 0
mirror of https://github.com/QuilibriumNetwork/ceremonyclient.git synced 2026-02-21 18:37:26 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity
29792d7f19
ceremonyclient/ARCHITECTURE.md
Vasyl Tretiakov 29792d7f19 update worker spawning details
2025-10-07 01:20:42 +03:00

22 KiB
Raw Blame History

Quilibrium Architecture

Overview

Quilibrium is a distributed protocol that leverages advanced cryptographic techniques including multi-party computation (MPC) for privacy-preserving compute. The system operates on a sharded network architecture where the master process runs global consensus while data worker processes run app-level consensus for their assigned shards, each maintaining their own hypergraph state, storage, and networking stack.

System Architecture

High-Level Architecture

┌─────────────────────────────────────────────────────────────────┐
│                       Client Applications                       │
│                  (CLI, RPC Clients, Applications)               │
└───────────────────────────────┬─────────────────────────────────┘
                                │
┌───────────────────────────────┴─────────────────────────────────┐
│                           RPC Layer                             │
│              (gRPC/REST APIs, IPCP2P Communication)             │
└───────────────────────────────┬─────────────────────────────────┘
                                │
┌───────────────────────────────┴─────────────────────────────────┐
│                       Master Process (Core 0)                   │
│   ┌──────────────┐  ┌─────────────┐  ┌─────────────────────┐    │
│   │    Global    │  │   Global    │  │    P2P Network      │    │
│   │  Consensus   │  │  Execution  │  │   (BlossomSub)      │    │
│   │   Engine     │  │   Engine    │  │                     │    │
│   └──────────────┘  └─────────────┘  └─────────────────────┘    │
│   ┌──────────────┐  ┌─────────────┐  ┌─────────────────────┐    │
│   │   Storage    │  │    Rust     │  │  Global Hypergraph  │    │
│   │  (PebbleDB)  │  │  Libraries  │  │       State         │    │
│   └──────────────┘  └─────────────┘  └─────────────────────┘    │
└───────────────────────────────┬─────────────────────────────────┘
                                │ IPCP2P
┌───────────────────────────────┴─────────────────────────────────┐
│                  Data Worker Process (Core 1)                   │
│   ┌──────────────┐  ┌─────────────┐  ┌─────────────────────┐    │
│   │     App      │  │Token/Hyper- │  │    P2P Network      │    │
│   │  Consensus   │  │graph/Compute│  │   (BlossomSub)      │    │
│   │   Engine     │  │  Execution  │  │                     │    │
│   └──────────────┘  └─────────────┘  └─────────────────────┘    │
│   ┌──────────────┐  ┌─────────────┐  ┌─────────────────────┐    │
│   │   Storage    │  │    Rust     │  │  Shard Hypergraph   │    │
│   │  (PebbleDB)  │  │  Libraries  │  │       State         │    │
│   └──────────────┘  └─────────────┘  └─────────────────────┘    │
└─────────────────────────────────────────────────────────────────┘
                                │
                    [Repeat for Core 2, 3, ...]

Core Components

1. Node Module (node/)

The main entry point containing the core system implementation with a multi-process architecture where each process has its own complete stack.

Process Types:

Master Process (Core 0)

Runs the global consensus and coordination:

  • Global Consensus Engine: System-wide consensus and coordination
  • Global Execution Engine: System-level operations
  • P2P Networking: Full BlossomSub stack for global communication
  • Storage (PebbleDB): Persistent storage for global state
  • Rust Library Bindings: Access to cryptographic operations
  • Global Hypergraph State: System-wide state management
Data Worker Processes (Cores 1+)

Each worker manages specific shards:

  • App Consensus Engine: Shard-level consensus
  • Execution Engines: Token, Hypergraph, and Compute engines
  • P2P Networking: Full BlossomSub stack for shard communication
  • Storage (PebbleDB): Persistent storage for shard state
  • Rust Library Bindings: Access to cryptographic operations
  • Shard Hypergraph State: Shard-specific state management

Application Layer (app/):

  • Node: Main node implementation (runs in master process)
  • DataWorkerNode: Worker node implementation (runs in worker processes)

Key Subsystems:

Consensus Engines (node/consensus/)

Two distinct consensus implementations:

Global Consensus (node/global/) - Master Process:

  • System-wide coordination
  • Cross-shard transaction ordering
  • Global state transitions
  • Network-wide fork choice

App Consensus (node/app/) - Worker Processes:

  • Shard-specific consensus
  • Local transaction ordering
  • Shard state transitions
  • Integration with global consensus

Shared Components:

  • DifficultyAdjuster: Dynamic difficulty adjustment
  • DynamicFeeManager: Dynamic Fee Market management
  • ProverRegistry: Enrolled prover registry
  • SignerRegistry: Block signer registry
  • Time Reels: VDF-based time consensus
Execution Engines (node/execution/)

Distribution across processes:

Global Execution Engine - Master Process:

  • System-wide operations
  • Cross-shard coordination
  • Global state updates

App Execution Engines - Worker Processes:

  • Token Engine: QUIL token operations within shards
  • Hypergraph Engine: CRDT-based state management for shards
  • Compute Engine: MPC-based privacy-preserving computation
P2P Networking (node/p2p/)

Each process (master and workers) has its own complete P2P stack:

  • BlossomSub: Custom pub/sub protocol
  • Peer Management: Discovery and connections
  • Public Channels: Point to point authenticated message routing
  • Private Channels: Onion-routing authenticated message channels
  • DHT Integration: Distributed peer discovery

Master process handles global bitmask, workers handle shard-specific bitmasks.

Storage Layer (node/store/)

Each process maintains its own PebbleDB instance:

  • Clock Store: Frame-based ordering
  • Coin Store: Locally managed token state
  • Data Proof Store: Cryptographic proofs
  • Hypergraph Store: Hypergraph state persistence
  • Peer Store: Network peer information

Storage is partitioned by process responsibility (global vs shard).

Hypergraph (hypergraph/)

Distributed graph structure:

  • Global Hypergraph: Maintained by master process
  • Shard Hypergraphs: Maintained by respective workers
  • CRDT Semantics: Conflict-free updates
  • Components:
    • Hypergraph: Core graph structure
    • Vertex: Node with associated data
    • Hyperedge: Multi-way relationships
Cryptographic Layer (types/crypto/)

Shared across all processes via Rust bindings (bls48581):

  • Proof Trees: Verkle tree structures
  • Channel Crypto: Secure E2EE communication

2. Client Module (client/)

Command-line interface for user interactions:

  • Node Management: Install, update, service control
  • Token Operations: Balance, transfer, mint operations
  • Hypergraph Operations: Deploy, store, update, query operations
  • Compute Operations: Deploy, execute, commit operations
  • Cross-Mint: Token bridging
  • Configuration: Network and RPC settings

3. Rust Crates (crates/)

High-performance cryptographic libraries used by all processes:

Core Cryptography:

  • bls48581: BLS signatures
  • bulletproofs: Transaction privacy via zero-knowledge proofs
  • vdf: Verifiable Delay Functions for time consensus
  • verenc: Verifiable encryption schemes

MPC Components:

  • ferret: Oblivious transfer protocols
  • channel: Secure communication protocols
  • rpm: Randomized Permutation Matrices

Supporting:

  • classgroup: Class group VDF implementation
  • WASM bindings for various libraries

4. Supporting Libraries

MPC and Privacy:

  • bedlam: Garbled circuit compiler/evaluator
  • emp-ot, emp-tool: Efficient MPC primitives
  • nekryptology: Additional cryptographic tools

Networking:

  • go-libp2p-blossomsub: Custom gossip for sharding
  • go-libp2p-kad-dht: DHT implementation
  • go-libp2p: Core P2P functionality

Storage:

  • node/pebble: High-performance key-value store
  • protobufs: Protocol definitions

5. Interfaces (types/)

Interfaces:

  • types/consensus/: Consensus-oriented interfaces:
    • AppConsensusEngine: The app shard level consensus engine
    • GlobalConsensusEngine: The global level consensus engine
    • DifficultyAdjuster: Manages the next difficulty following the prior frame
    • EventDistributor: The central nervous system to internal event control flows, connects time reel events, consensus engine events, and emits ordered control events. Manages halts, resumes, and state rewind/replay events.
    • DynamicFeeManager: Tracks and calculates dynamic fee market multipliers
    • RewardIssuance: The issuance policy of the network
    • ProverRegistry: Tracks prover information from ongoing hypergraph state transitions
    • SignerRegistry: Tracks general signer information, including private messaging channel keys
    • AppFrameValidator: Validator path for app shard frames
    • GlobalFrameValidator: Validator path for global frames
  • types/crypto/:
    • BlsAggregateOutput: Methods for handling aggregate signatures
    • BlsKeygenOutput: Methods for handling BLS keygen outputs
    • BulletproofProver: Provides construction and verification of outputs related to bulletproofs
    • FrameProver: Methods for generating proofs on frames
    • InclusionProver: Methods for generating KZG proofs
    • Multiproof: Methods for multiproofs
    • BlsConstructor: Methods for creating BLS keys
    • DecafConstructor: Methods for creating Decaf 448 keys
    • DecafAgreement: Methods specific to key agreement with Decaf 448 keys
    • Signer: Generic signer interface to unify all managed key types
    • Agreement: Generic key agreement interface to unify all DH agreed keys
    • VerifiableEncryptor: Methods for producing verifiable encryption proofs
    • VerEncProof: Methods for handling the proof outputs
    • VerEnc: Methods for handling the compressed proof outputs for storage
  • types/execution/intrinsics/: Intrinsics and intrinsic operation interfaces:
    • Intrinsic: Methods for managing intrinsics
    • IntrinsicOperation: Methods for performing operations of the intrinsics
  • types/execution/state/: State management related interfaces:
    • State: Methods for initializing, modifying, and deleting state, and retrieving state transition outputs
  • types/execution/: Execution engine encapsulation
    • ShardExecutionEngine: Shard-specific execution container methods
  • types/hypergraph/: Hypergraph-oriented interfaces
  • types/keys/: Key management interfaces
  • types/p2p/: PubSub related interfaces
  • types/store/: KVDB store related interfaces

Mock Implementations:

  • MockBlsConstructor: BLS Key Constructor, used by key manager
  • MockBLSSigner: BLS Keypair, used in many places
  • MockBulletproofProver: Bulletproof prover, used by token intrinsic
  • MockDecafConstructor: Decaf 448 Key Constructor, used by key manager and token intrinsic
  • MockVertex, MockHyperedge, MockHypergraph: Hypergraph components
  • MockInclusionProver: KZG prover
  • MockMultiproof: Multiproof output
  • MockKeyManager: Key manager, used everywhere
  • MockPubSub: PubSub-based message system, used by consensus engines
  • MockVerifiableEncryptor, MockVerEncProof, MockVerEnc: Verifiable encryption, used in intrinsics, hypergraph storage

Tries: All trie related constructions are contained within types for easy WASM generation.

Architectural Patterns

1. Shared-Nothing Process Architecture

Each process is self-contained with its own:

  • Consensus engine (global or app)
  • Execution engine(s)
  • P2P networking stack
  • Storage layer
  • Cryptographic libraries

Benefits:

  • Process independence
  • Failure isolation
  • Horizontal scalability

2. Sharded State Management

State is partitioned across processes:

  • Master process: Global state and coordination
  • Worker processes: Shard-specific state
  • CRDT hypergraph for convergence

3. Multi-Party Computation (MPC)

Privacy-preserving computation:

  • Garbled circuits via bedlam
  • Oblivious transfer for private inputs
  • General-purpose privacy beyond just transactions

4. Layered Consensus

Two-tier consensus system:

  • Global consensus for system-wide coordination
  • App consensus for shard-level operations
  • Time Reel VDF for temporal ordering

5. IPC Communication

Inter-process communication design:

  • Structured message passing
  • Minimal shared state

Security Architecture

1. Privacy Layers

Transaction Privacy:

  • Bulletproofs for confidential transactions
  • Zero-knowledge range proofs and sum checks

Computation Privacy:

  • MPC for general computations
  • Garbled circuits for function privacy
  • Oblivious transfer for input privacy

Data Privacy:

  • Verifiable encryption for settled state

2. Process Isolation

  • Independent process spaces
  • Limited IPC surface
  • Resource isolation per process

3. Network Security

  • Peer reputation scoring
  • Sybil resistance
  • Eclipse attack prevention
  • Shard-aware routing

4. Binary Verification

  • Multi-signature verification
  • Threshold signing schemes
  • Automated signature updates

System Boundaries

Process Boundaries

  • Master process: Global operations
  • Worker processes: Shard operations
  • Clear IPC interfaces between processes

Consensus Boundaries

  • Global consensus: Cross-shard coordination
  • App consensus: Intra-shard operations
  • Well-defined interaction protocols

State Boundaries

  • Global state in master process
  • Shard state in worker processes
  • CRDT merge protocols for synchronization

Network Boundaries

  • Each process has independent P2P stack
  • Topic-based message routing
  • Shard-aware peer connections

Build and Deployment

Build System

  • Hybrid Go/Rust via CGO
  • Platform-specific optimizations
  • Static linking for distribution

Configuration

  • Per-process configuration
  • Environment variable support
  • YAML configuration files

Monitoring

  • Prometheus metrics per process
  • Grafana dashboards
  • Structured logging

Design Rationale

1. Independent Process Stacks

  • Decision: Each process has full stack
  • Rationale: Maximum isolation, independent scaling, failure resilience

2. Two-Tier Consensus

  • Decision: Separate global and app consensus
  • Rationale: Scalability through sharding while maintaining global coordination

3. MPC for General Privacy

  • Decision: MPC beyond just transactions
  • Rationale: Flexible privacy for arbitrary computations

4. CRDT Hypergraph

  • Decision: CRDT-based state model
  • Rationale: Natural convergence in distributed sharded system

5. Rust Cryptography

  • Decision: Rust for crypto operations
  • Rationale: Performance and memory safety for critical operations

Development Guidelines

Architecture Principles

  • Maintain process independence
  • Respect consensus boundaries
  • Minimize cross-shard communication
  • Preserve privacy guarantees

Code Organization

  • Clear separation by process type
  • Shared interfaces and types
  • Process-specific implementations

Testing Strategy

  • Unit tests per component
  • Integration tests per process
  • Cross-process interaction tests
  • MPC protocol verification

Frequently Asked Questions (FAQ) with Code References

Getting Started

Q: Where is the main entry point?

  • Main entry: node/main.go
  • The main() function handles initialization, and manages the lifecycle

Q: How does the multi-process architecture work?

  • Process spawning: node/worker/manager.go (see spawnDataWorkers function)
  • Master process runs on core 0, creates and starts Global Consensus Engine with embedded Worker Manager
  • Upon start, Worker Manager spawns workers for cores 1+ using exec.Command
  • Each worker receives --core parameter to identify its role

Q: How do I run the node?

  • Build: ./node/build.sh (creates static binary)
  • Run: ./node/node (starts master process which spawns workers)
  • Configuration: config.yml or environment variables

Architecture & Design

Q: How do processes communicate (IPC)?

  • Uses IPC over private pubsub topics for structured message passing
  • Master process acts as coordinator, workers connect via IPC

Q: Where is consensus implemented?

  • App Consensus (workers): node/consensus/app/app_consensus_engine.go
  • Factory: node/consensus/app/factory.go
  • Consensus interface: node/consensus/consensus_engine.go

Q: Where is the hypergraph implementation?

  • Main interface: hypergraph/hypergraph.go
  • Core components:
    • Vertex: hypergraph/vertex.go
    • Hyperedge: hypergraph/hyperedge.go
    • Atoms: hypergraph/atom.go
    • Proofs: hypergraph/proofs.go

Networking & P2P

Q: How is P2P networking initialized?

  • BlossomSub: node/p2p/blossomsub.go (NewBlossomSub)
  • Worker P2P ports: Base port + worker index
  • Each process has independent P2P stack

Q: How does cross-shard communication work?

  • Through master process coordination via P2P
  • Topic-based routing in BlossomSub
  • Shard-aware peer connections
  • Majority of cross-shard communication requires only proof data, keeping shards conflict-free

Execution & State

Q: Where are execution engines created?

  • Factory: node/execution/engines/factory.go (CreateExecutionEngine)
  • Engine types: Global, Compute, Token, Hypergraph
  • Batch creation: CreateAllEngines

Q: How is state stored?

  • PebbleDB wrapper: node/store/pebble.go
  • Clock store: node/store/clock.go
  • Localized Coin store: node/store/coin.go
  • Each process has its own DB instance

Cryptography & Privacy

Q: Where is MPC/garbled circuits used?

  • Bedlam compiler: bedlam/ directory
  • Compute intrinsic: node/execution/intrinsics/compute/compute_deploy.go
  • Client deployment initializes the garbler

Q: Where are the cryptographic libraries?

  • Rust crates: crates/ directory
    • BLS: crates/bls48581/
    • Bulletproofs: crates/bulletproofs/
    • VDF: crates/vdf/
  • Go bindings: Generated via CGO in each crate

Q: How are bulletproofs integrated?

  • Rust implementation: crates/bulletproofs/
  • Used for transaction privacy (amount hiding)
  • Go bindings in bulletproofs/ directory

Configuration & Operations

Q: How do I configure the node?

  • Main config: config/config.go (Config struct)
  • Engine config: config/engine.go
  • Data worker config
  • Load with YAML file or environment variables

Q: What are the RPC/API endpoints?

  • Node RPC: node/rpc/node_rpc_server.go
  • gRPC definitions: protobufs/node.proto

Q: How do I monitor the node?

  • Prometheus metrics throughout codebase
  • Grafana dashboards: dashboards/grafana/
  • Structured logging with configurable levels

Development

Q: How is the project built?

  • Node build: node/build.sh (static linking, platform detection)
  • Client build: client/build.sh
  • Rust integration: Various generate.sh scripts
  • Docker: Multiple Dockerfiles for different scenarios

Q: What happens when a worker crashes?

  • Master process monitors worker processes
  • Automatic restart logic in spawnDataWorkers
  • Process isolation prevents cascade failures

Q: How do I add a new execution engine?

  • Implement the interface in node/execution/
  • Add to factory in node/execution/engines/factory.go
  • Register intrinsics as needed

Common Code Paths

Transaction Flow:

  1. RPC endpoint receives transaction
  2. Routed to appropriate shard via worker
  3. App consensus orders transaction
  4. Execution engine processes
  5. State updated in hypergraph
  6. Proof generated and stored

Consensus Flow:

  1. Time reel provides ordering
  2. App consensus in workers for shards
  3. Global consensus coordinates cross-shard
  4. Fork choice based on VDF proofs

P2P Message Flow:

  1. BlossomSub receives message
  2. Topic routing to appropriate handler
  3. Shard-specific messages to workers
  4. Global messages to master process