Following the 2024 roadmap, the Broxus team has conducted research, to help find a solution for one of Everscale's primary technical challenges: achieving fast finality as the workload increases. In particular, developers touched on that topic during the panel discussion at the latest Everpoint conference.
As a practical outcome of this research, experts undertook the task of creating a protocol that could leverage the strengths of TVM-based networks while integrating the latest advancements in blockchain technologies.
Major Flaws in the TVM Networks’ Consensus Algorithm
Although designed as the fastest to date, TVM blockchains have some major hurdles, preventing them from achieving fast finality.
The first issue is transaction finality. The current architecture of cross-shard interaction is structured through master blocks, and the time to finalize a transaction chain depends on the number of accounts and their distribution across shards.
The second hindrance is performance. The current block production in a TVM network works as follows:
- A collator generates a block and sends it to validators, with a strict timing requirement for the collator to complete block collation.
- Validators receive the block, verify it, and sign it.
The bulk of the transaction execution workload falls on the collator. The process of creating a block can halt once certain limits are reached, such as time spent, block size, or changes in the block’s logical time. Essentially, much of the time within the consensus is spent moving blocks across the network for validation rather than executing transactions.
Finally, there is a security issue. Currently, shard blocks are produced by a randomly selected subset of validators. To enhance security, it is necessary to increase the number of validators or introduce a role of verifiers.
The Suggested Protocol’s Architecture
Within the research, technical experts specifically considered modern DAG-based protocols. In one peer-reviewed paper they discovered, the authors proposed the DAG protocol called Mysticeti, which, according to the publication, can achieve up to 400,000 transactions per second (TPS).
Based on the findings, a model was proposed for the protocol, built around three core components:
- DAG-based consensus algorithm for ordering and processing external messages.
- Multithreaded collator.
- Validator.
Let’s break down each component and its benefits:
- Consensus. Validators receive external messages from users and publish them into the consensus protocol. Upon reaching the consensus, all nodes obtain an ordered, agreed-upon queue of external messages. This queue is then filtered by the validator to remove duplicates before being passed to the collator.
- Collator. The collator is responsible for executing both internal and external messages, producing blocks, and altering the blockchain state. To perform its tasks, the collator requires the queue of external messages from the consensus and the current blockchain state. Unlike in other systems, the collator does not need to synchronize with other nodes, allowing it to continuously produce blocks and fully utilize the server’s computational power. These blocks are then sent for validation.
- Validator. The validator’s job is to verify that the block’s hash matches that of at least two-thirds of the network’s nodes. Validators exchange only block hashes and signatures, rather than the blocks themselves, significantly reducing network traffic. A node needs the validation to ensure it reaches the same result as other nodes in the network. Once signatures are collected, the node can distribute the signed blocks to network clients.
This novel approach to consensus and block production forms the foundation of impressive performance and scalability, with the suggested protocol’s design having a potential to become a major technical leap for TVM blockchains.
