Parallel Consensus

Scalaris leverages DAG-based leaderless consensus protocols such as Narwhal & Bullshark and Mysticeti to implement a novel approach known as parallel consensus. This approach fundamentally enhances the efficiency and scalability of the consensus process by eliminating traditional bottlenecks and communication overhead.

Key Features of Parallel Consensus

1. Elimination of Communication Overhead:

• Traditional consensus protocols often require multiple rounds of communication among validators to agree on the same set of transactions. This process can be slow and resource-intensive. Scalaris eliminates this overhead by allowing validators to independently examine their local view of the DAG and fully order all vertices without needing additional message exchanges. This leaderless approach drastically reduces the communication overhead typically associated with consensus processes.

1. Separation of Data Propagation and Consensus:

• Scalaris separates the tasks of data propagation and consensus, allowing these processes to run in parallel. Transaction data is efficiently disseminated throughout the network, while the consensus protocol focuses on ordering metadata references rather than the full transaction data. This separation enhances throughput and simplifies the consensus process, as validators can commit new blocks by referencing the local DAG structure.

1. Local View Consensus:

• Each validator in Scalaris can independently reach consensus by examining its local view of the DAG. Validators make commit decisions based on the local structure of the DAG, which includes all necessary information about previous transactions and blocks. This approach ensures that consensus can be achieved quickly and efficiently without the need for extensive coordination among validators.