Skip to main content

Savitri Network Technical Documentation

Welcome to the Savitri Network Technical Documentation. This is not a high-level overview; it is a deep dive into the engineering of a non-EVM, state-sharded Directed Acyclic Graph (DAG) built entirely in Rust.

Savitri is architected to solve the scalability trilemma by shifting the burden of validation from centralized server clusters to a decentralized, hardware-aware mobile network.

Quick Navigation: Where to Start?​

To streamline your integration and review process, we have categorized our documentation by engineering domains:

  • Core Architecture: Deep dive into our non-EVM DAG structure and state-sharding logic.
  • Consensus (PoU): Technical breakdown of Proof-of-Utility scoring (Uptime, Latency, Integrity, and Reputation).
  • Deterministic Execution: Understanding our fixed-point arithmetic layer and cross-platform (ARM/x86) consistency.
  • P2P Networking: Analysis of our NUMA-aware signature verification and adaptive gossip protocol.

Core Engineering Principles​

Every commit to Savitri is governed by three fundamental requirements:

  1. Bit-Perfect Determinism
    No floating-point operations in the execution layer. We ensure identical state transitions across heterogeneous hardware using fixed-point math.

  2. Memory Safety
    Zero-cost abstractions and strict ownership via Rust. We eliminate data races and undefined behavior (UB) by design.

  3. Hardware Awareness
    The network optimizes itself for the underlying CPU topology. Our nodes utilize idle cycles and NUMA-aware scheduling to maximize throughput without compromising device longevity.

The Goal: 45k+ TPS on the Edge​

Traditional blockchains sacrifice decentralization for speed by requiring $10k+ servers. Savitri reverses this. By optimizing the serialization overhead (Bincode) and implementing adaptive batching, we enable high-throughput validation on the devices people already own.