Network Architecture

Heterogeneous Node Architecture — Right Node, Right Workload

Six specialized node types, each optimized for a specific workload. Validation nodes run consensus. FHE nodes crunch encrypted computations on GPUs. Web nodes serve on-chain websites. API nodes bridge external systems. Batch nodes process millions of operations. Storage nodes hold distributed blobs. One network, six specializations, maximum performance.

Specialized Infrastructure

Six Node Types

Unlike homogeneous networks where every node does everything, ATSHI assigns the right hardware and software profile to each workload class

  • Validation Nodes — Core Consensus: the backbone of the network. These nodes run ARCH consensus, validate transactions, maintain the canonical chain state, and enforce protocol rules. Optimized for CPU efficiency and low-latency networking. Every transaction passes through validation nodes before becoming final.
  • FHE Compute Nodes — GPU-Accelerated: dedicated to Fully Homomorphic Encryption workloads. Equipped with high-end GPUs (NVIDIA A100/H100 class) that accelerate FHE operations by orders of magnitude. These nodes process encrypted computations — confidential DeFi, private ML inference, sealed-bid auctions — without ever seeing plaintext data.
  • Web Hosting Nodes — On-Chain Sites: serve websites and dApps stored on the blockchain. Optimized for HTTP/HTTPS throughput with integrated SSL/TLS termination, caching, and GeoDNS routing. These nodes turn the blockchain into a global CDN — content-addressed, verifiable, and censorship-resistant.
  • API Gateway Nodes — REST/GraphQL Interface: the bridge between the blockchain and external systems. Expose ATSHI data and operations through familiar REST and GraphQL APIs. Rate limiting, authentication, request transformation, and response caching are built in. Enterprise systems connect to ATSHI through API nodes without needing blockchain-specific code.
  • Batch Processing Nodes — Bulk Operations: designed for high-throughput, high-volume workloads like mass token distributions, data migrations, bulk attestations, and IoT data ingestion. Batch nodes aggregate operations, generate Merkle proofs for the batch, and settle on-chain with a single root hash — achieving millions of operations per settlement.
  • Storage Nodes — Distributed Blob/File Storage: provide persistent, distributed storage for large files, media, documents, and dataset blobs. Content-addressed with erasure coding for durability. Storage nodes handle the heavy lifting of data persistence while keeping the main chain lean and fast.

Node Type Map

⚙ Validation Core consensus, transaction validation, chain state
💻 FHE Compute GPU-accelerated encrypted computation
🌐 Web Hosting On-chain websites, SSL, GeoDNS
🔌 API Gateway REST/GraphQL, rate limiting, auth
⚡ Batch Processing Bulk ops, Merkle proofs, mass settlement
🗃 Storage Distributed blobs, erasure coding, files
6 specialized types — each optimized for its workload class
Consensus Validation nodes — CPU-optimized, low-latency
Compute FHE nodes — GPU-accelerated (A100/H100)
Serving Web + API nodes — HTTP/GraphQL optimized
Throughput Batch nodes — millions of ops per settlement
High-Volume Operations

Batch Processing Pipeline

From ingestion to on-chain settlement — how batch nodes process millions of operations with a single Merkle proof

  • Ingest: batch nodes accept high-volume operation streams via dedicated ingestion endpoints. IoT sensor data, mass token distributions, bulk attestations, data migrations — operations arrive at rates traditional blockchains cannot handle. Ingestion buffers absorb burst traffic and queue operations for processing.
  • Process: operations are validated, transformed, and organized into batches. Each batch groups related operations for efficient processing. Invalid operations are rejected with detailed error reports. Business logic executes against the current state, producing a set of state transitions ready for proof generation.
  • Prove: a Merkle tree is constructed over all state transitions in the batch. The Merkle root serves as a cryptographic commitment to the entire batch — any single operation can be independently verified against the root without replaying the entire batch. Compact proofs, massive throughput.
  • Settle: the Merkle root and batch metadata are submitted to the main chain via a single transaction. This one transaction settles the entire batch — potentially millions of individual operations. Validation nodes verify the root, and the batch becomes part of the immutable chain history. Individual operations can be verified on demand using Merkle inclusion proofs.

Batch Pipeline

📥 Ingest
→
⚙ Process
→
📜 Prove
→
✅ Settle
Millions of operations → one Merkle root → one on-chain transaction

Performance by Node Type

Validation
Fast
FHE GPU
100x CPU
Web
Global CDN
API
REST/GQL
Batch
Millions/tx
Storage
Erasure-coded
Why Heterogeneous Nodes

Advantages

Homogeneous networks force every node to do everything. ATSHI lets each node do what it does best — delivering specialized performance, flexible deployment, and enterprise-grade workload support.

🚀

Specialized Performance

Each node type is optimized for its workload. FHE nodes have GPUs. Web nodes have caching. Batch nodes have high-throughput ingestion. No compromise hardware, no bottleneck from mixed workloads. Every operation runs on purpose-built infrastructure.

Optimized · Purpose-Built
🔧

Flexible Deployment

Run only the node types you need. An enterprise might run validation + API + storage nodes for a private network. A DeFi protocol adds FHE compute nodes. A content platform adds web hosting nodes. Scale each type independently based on demand.

Modular · Independent Scaling
🏢

Enterprise Workloads

Batch processing handles mass operations that would overwhelm traditional blockchains. API gateways provide familiar REST/GraphQL interfaces. Storage nodes handle large files. Enterprise IT teams deploy ATSHI without rethinking their architecture.

Batch · API · Storage
💻

GPU Acceleration

FHE operations that take minutes on CPU execute in seconds on GPU. NVIDIA A100/H100 acceleration makes confidential computing practical for real-time applications — encrypted DeFi trades, private ML inference, sealed-bid auctions at production speed.

A100/H100 · 100x Faster
📈

Batch Throughput

Millions of operations settle with a single Merkle root transaction. IoT data streams, mass distributions, bulk attestations — workloads that are economically impossible on per-transaction chains become trivial with batch processing nodes.

Millions/Tx · Merkle Proofs
🌎

Global Distribution

Web hosting and API gateway nodes deploy worldwide with GeoDNS routing. Storage nodes use erasure coding for durability across regions. The network adapts to geographic demand, not geographic limitations.

GeoDNS · Erasure-Coded · Global
Network Function

Network & Geo Pathing

Geographically aware node organization for optimized replication, routing, and sovereignty enforcement

  • Geo-patch organization — the ATSHI network is geographically aware. Nodes are organized into geo-patches — clusters of nodes grouped by physical location and region. This structure enables optimized replication and intelligent routing across the global network.
  • Geographic replication — transaction chains are replicated across multiple geographic zones for resilience. Data is never concentrated in a single region, ensuring availability even during regional outages or network partitions.
  • Nearest-patch routing — the network automatically routes requests to the nearest geo-patch for minimal latency. Whether a user is in Paris, Tokyo, or São Paulo, their requests reach the closest available nodes without manual configuration.
  • Geographic sovereignty enforcement — data can be constrained to specified regions, ensuring compliance with data residency requirements (GDPR, local regulations). The geo-patch architecture makes geographic sovereignty a protocol-level guarantee, not an operational policy.
  • Optimized content delivery — for on-chain hosting, geo-pathing enables CDN-like content delivery where websites and assets are served from the geographically nearest nodes, dramatically reducing load times worldwide.
  • Data residency compliance — enterprises can specify that certain transaction chains must remain within specific geographic boundaries. The network enforces these constraints at the replication level, ensuring data never leaves permitted jurisdictions.

Geo-Patch Architecture

Geo-Patches Nodes organized by geographic region for optimized routing
Replication Transaction chains replicated across geographic zones for resilience
Nearest Routing Requests automatically routed to closest geo-patch for minimal latency
Sovereignty Data constrained to specified regions — geographic enforcement at protocol level
CDN Delivery On-chain hosting served from nearest nodes worldwide
Compliance Data residency requirements enforced at the replication level
How We Compare

Node Architectures Compared

Homogeneous node networks force the same hardware requirements on every participant. ATSHI's heterogeneous architecture matches hardware to workload for maximum efficiency.

Feature Ethereum Solana Hyperledger Fabric Cosmos ATSHI Network
Node Model Homogeneous Homogeneous (high-spec) Peer + Orderer Homogeneous 6 specialized types
Hardware Requirements Moderate (all nodes) Very high (all nodes) Moderate Moderate Per-type optimized
GPU Compute No No No No FHE nodes with A100/H100
Built-in Web Hosting No No No No Dedicated web hosting nodes
API Gateway External (Infura, Alchemy) External (RPC providers) Peer SDK External RPC Native REST/GraphQL nodes
Batch Processing No native (L2 rollups) No Limited No Dedicated batch nodes + Merkle settlement
File Storage External (IPFS/Arweave) External Off-chain (CouchDB) External Native storage nodes, erasure-coded
Independent Scaling No (all nodes equal) No (all nodes equal) Limited (peer vs orderer) No Each type scales independently
Enterprise Integration External tools needed External tools needed Good (SDK) External tools needed Native API + Batch + Storage

Run the Right Node for Your Workload

Stop forcing every workload through a one-size-fits-all node. ATSHI's heterogeneous architecture gives you GPU-accelerated FHE computation, native web hosting, REST/GraphQL API gateways, batch processing that settles millions of operations in a single transaction, and distributed file storage — all as first-class network participants. Run validation nodes for consensus. Add FHE nodes for confidential compute. Deploy web nodes for on-chain hosting. Scale each type independently. The right node for the right workload.

Explore Full Technology Stack Invest in ATSHI