Exploring a new path to blockchain scalability through a DAG-based architecture

Blockchain

Exploring a new path to blockchain scalability through a DAG-based architecture

Can a blockchain scale without breaking the promise of decentralization？ Waterfall Network argues yes by leaving the linear blockchain model behind. In July 2024, Waterfall Network launched its mainnet and joined the short list of layer-1 protocols that depart from the traditional blockchain architecture. In its latest report, Cointelegraph Research investigates Waterfall Networks architecture and validator design to assess whether it can become a sustainable, alternative base layer.

The full Cointelegraph Research report on Waterfall includes validator roles, DAG propagation mechanics and ecosystem overview. Download the full report here

DAG Architecture and Why It Matters for Blockchain Scaling

Waterfall Network is a layer-1 protocol that implements a Directed Acyclic Graph (DAG) as its foundational ledger. A DAG is a non-linear data structure that arranges blocks as vertices in a directed graph.

Each new block or transaction references one or more prior blocks. This eliminates the requirement for global sequencing at the time of transaction creation. Instead of a single ordered chain, the ledger evolves as a concurrent graph of interlinked blocks, growing in many directions.

Unlike earlier DAG systems, Waterfall does not entirely discard linear ordering. Instead, it separates concerns across two subsystems: the Shard Network and the Coordinating Network.

The Shard Network has the DAG structure and scales horizontally using hierarchical fractal sharding. Each shard can split into smaller subshards as demand grows and transactions can be handled independently. However, all shards ultimately sync to a shared global state. The coordinating network finalizes transactions from the DAG by selecting spine blocks and putting them in sequence to ensure network-wide consistency. This allows Waterfall to process transactions in parallel while maintaining a global state without overloading any single part of the network.

Waterfalls Validator Design and TPS

During benchmarking by Chainspect, Waterfall reached a peak throughput of 12,777 TPS. This outperforms high-speed EVM-based competitors such as Monad and SEI. Its validator framework supports up to 1.5 million participants. Running a validator also only requires 2-core CPUs and 8GB RAM, far below the node requirements for most high-performance chains. This lets lightweight hardware join the network and avoids centralization.