Storb is a decentralized storage protocol integrated into the Bittensor ecosystem as Subnet 26, designed to provide a scalable and fault-tolerant solution for object storage. The “product” is essentially a distributed storage network where users can upload files, and the system handles the rest—splitting, distributing, and storing the data across a network of miners while ensuring it can be retrieved efficiently. When a user uploads a file, validators take the lead: they break the file into smaller chunks using erasure coding, a method that adds redundant pieces so the file can be reconstructed even if some chunks are missing, and then distribute these chunks to miners for storage. Miners, in turn, store these pieces and are periodically challenged by validators to prove they still hold the data, ensuring accountability. The retrieval process is just as seamless—validators use a Distributed Hash Table (DHT) to track where each chunk is stored, allowing the system to reassemble the file when requested. Storb’s design prioritizes reliability and accessibility, with features like:

• Erasure-Coded Redundancy: Ensures data can be recovered even if some miners go offline.
• Incentivized Participation: Miners are rewarded with TAO for storing data and responding to challenges.
• Global Distribution: Data is spread across a worldwide network, eliminating single points of failure.

Unlike centralized storage providers that rely on a few data centers, Storb leverages the collective power of its miners to create a system that’s both resilient and decentralized, making it a pioneering solution for Web3 applications that need secure, distributed storage.

Technical Architecture

Storb’s technical architecture is a carefully engineered system that balances efficiency, scalability, and security, integrating seamlessly with the Bittensor network while handling the complexities of decentralized storage. At its core, Storb operates as a subnet within Bittensor, consisting of up to 256 nodes (neurons), with 64 slots reserved for validators and the rest for miners, following Bittensor’s standard subnet structure. The process begins with validators, who handle file uploads by splitting files into smaller pieces using erasure coding—specifically, a Reed-Solomon encoding scheme that adds redundant data to ensure recoverability. These pieces are then distributed to miners, who store them on their local systems. To track the location of these chunks, Storb uses a Distributed Hash Table (DHT), a decentralized system that maps file pieces to their respective miners, enabling efficient retrieval. The architecture also includes a robust scoring system to evaluate miner performance, based on three key metrics:

• Latency (20%): Measures how quickly miners respond to requests, ensuring speed is prioritized.
• Success Rate (30%): Assesses the reliability of storage and retrieval, rewarding miners who consistently deliver.
• PDP Challenge (50%): Uses Proof of Data Possession (PDP) challenges, where miners must provide cryptographic proofs that they still hold the data, verifying integrity without transferring the entire file.

Storb leverages SQLite as its backend for storing metadata, providing a lightweight and efficient way to manage file information. While storage and retrieval operations happen off-chain for speed, the incentive mechanism is tied to the Bittensor blockchain, where validators set weights for miners based on their scores, and TAO rewards are distributed accordingly. This hybrid approach—off-chain storage with on-chain incentives—ensures Storb is both fast and future-ready, capable of scaling as the Bittensor network grows.