Skip to main content

Technical Specification Overview

System Architecture​

Savitri Network is an enterprise-grade Layer 1 blockchain implementing a custom execution environment with BFT consensus, SIMD-optimized transaction processing, and adaptive weight scheduling.

Core Components​

Consensus Layer

  • src/consensus/ - BFT consensus engine with Proof of Unity (PoU) scoring
  • src/core/slot_scheduler.rs - Deterministic slot assignment and leader election
  • Committee size: 4-256 validators with bond requirements (1M tokens minimum)

Execution Layer

  • src/executor/ - Transaction execution pipeline with SIMD optimization
  • src/executor/dispatcher.rs - ExecutionDispatcher with adaptive weights
  • src/executor/score_cache.rs - Thread-safe caching for cross-batch optimization

Storage Layer

  • src/storage/ - RocksDB-based persistence with column families
  • src/storage/monolith.rs - Monolith block aggregation with ZKP proofs
  • Column families: CF_ACCOUNTS, CF_TRANSACTIONS, CF_BLOCKS, CF_CONSENSUS, CF_BONDS

Network Layer

  • src/p2p/ - libp2p-based networking with gossipsub message propagation
  • src/p2p/messages.rs - Protocol message definitions and serialization
  • src/networking/ - Connection management and peer discovery

Cryptographic Layer

  • src/crypto/ - Ed25519 signatures, Blake3 hashing, curve25519 operations
  • src/zkp/ - Zero-knowledge proof system for monolith verification
  • SIMD-enabled cryptographic operations with runtime feature detection

Module Dependencies​

consensus → storage → crypto
    ↓           ↓        ↓
executor ← mempool ← p2p
    ↓
contracts ← governance

Key Architectural Decisions​

1. Custom Execution Environment

  • Non-EVM compatible virtual machine with custom bytecode
  • Function selector: keccak256(function_signature)[0:4]
  • Storage layout: Reserved slots 0-99 for BaseContract, custom slots 100+

2. SIMD Optimization

  • Transaction scoring vectorization using AVX2/NEON intrinsics
  • Runtime feature detection with automatic scalar fallback
  • 2-3x theoretical speedup on x86_64, 1.5-2x on ARM

3. Adaptive Weight Scheduling

  • Dynamic fee-to-block-space ratio adjustment based on mempool conditions
  • Feedback mechanism with configurable update intervals
  • Cross-batch optimization through score caching

4. Monolith Architecture

  • Block aggregation with cryptographic commitment to headers
  • Zero-knowledge proof verification for state consistency
  • Light node optimized verification without full state download

Performance Characteristics​

Throughput Targets

  • TPS: 10,000+ with SIMD optimization
  • Block time: 500ms target
  • Finality: 2-3 blocks (1-1.5 seconds)

Resource Requirements

  • Full node: 16GB RAM, 4+ CPU cores, 1TB SSD
  • Validator: 32GB RAM, 8+ CPU cores, 2TB NVMe SSD
  • Light node: 512MB RAM, mobile CPU optimized

Network Specifications

  • P2P protocol: libp2p with gossipsub
  • Message serialization: bincode with Blake3 integrity checks
  • Peer discovery: mDNS + bootstrap nodes
  • Maximum peers: 50 per node

Security Model​

Consensus Safety

  • BFT with f=⌊(n−1)/3⌋f = \lfloor(n-1)/3\rfloor fault tolerance
  • Slashing: 50% bond penalty for equivocation
  • Finality: Cryptographic certificates with 2f+12f+1 validator signatures

Cryptographic Guarantees

  • Signature scheme: Ed25519 with batch verification
  • Hash functions: Blake3 for performance, SHA3 for compatibility
  • ZKP system: PLONK-based for monolith verification

Network Security

  • TLS encryption for all P2P connections
  • Peer reputation scoring with automatic misbehavior detection
  • Rate limiting and DoS protection at protocol level

Implementation Status​

Completed Components

  • Core consensus engine with PoU scoring
  • SIMD-optimized transaction dispatcher
  • RocksDB storage layer with monolith support
  • libp2p networking with message propagation
  • Basic smart contract platform

In Progress

  • Advanced governance system
  • Cross-chain bridge implementation
  • Sharding architecture
  • Mobile wallet SDK

Known Limitations

  • EVM compatibility requires bridge implementation
  • Mobile validator participation not supported (bond requirements)
  • Cross-shard transaction coordination under development

Validation Criteria​

Determinism Requirements

  • All transaction execution must be deterministic across architectures
  • SIMD vs scalar computation divergence < 10−1010^{-10}
  • State root computation must be cryptographically verifiable

Performance Benchmarks

  • SIMD speedup: ≥1.25x for batches ≥32 transactions
  • Cache hit rate: ≥80% for typical workloads
  • Network latency: ≤100ms for intra-region propagation

Security Guarantees

  • Zero knowledge proof verification for all monoliths
  • BFT consensus safety under ff Byzantine faults
  • Cryptographic integrity for all stored data

This specification serves as the foundation for all implementation decisions and audit requirements.