Smart Red Belly Blockchain: Enhanced Transaction Management for Decentralized Applications

TL;DR

This paper tackles the bottlenecks of transaction management in Ethereum-like blockchains that hinder DApp performance. It proposes SRBB, a VM optimized with a novel transaction validation reduction and per sub-block processing, integrated with a fast consensus to form Smart Red Belly Blockchain. The approach yields substantial gains, achieving up to 4000 TPS peak and about 2000 TPS on geo-distributed deployments, while outperforming six state-of-the-art blockchains on a Nasdaq-based Diablo workload. The work demonstrates that careful coordination of validation, storage, and execution can dramatically improve throughput and latency for DApp-rich blockchains without compromising safety or finality, enabling scalable Web3 infrastructure across continents.

Abstract

Decentralized Applications (DApps) have seen widespread use in the recent past driving the world towards a new decentralized version of the web known as Web3.0. DApp-supported blockchains like Ethereum have largely been responsible for this drive supporting the largest eco-system of DApps. Although the low performance provided by Ethereum has been a major impediment to realizing a decentralized web, several high-performance blockchains have been introduced recently to bridge this gap. Most of these blockchains rely on consensus optimizations. Only a few enhance other parts of the blockchain protocol that involves transaction management: the validation of transactions, broadcast of transactions, encapsulation and dissemination of blocks with transactions, re-validation and execution of transactions in blocks, storage of blocks, and confirmation of transaction commits to senders upon request. In this paper, we enhance transaction management by introducing a novel transaction validation reduction and a per sub-block processing to optimize the block storage. We empirically show the performance improvements gained by our enhanced transaction management in the Smart Red Belly Blockchain (SRBB) VM we develop. Finally, we integrate our SRBB VM to an already optimized consensus from a known blockchain to develop the Smart Red Belly Blockchain. Our results show that SRBB achieves a peak throughput of 4000 TPS and an average throughput of 2000 TPS on 200 nodes spread across 5 continents. SRBB outperforms 6 other blockchains when running the exchange DApp featuring a real workload trace taken from Nasdaq.

Figures (5)

• Figure 1: The architecture of a SRBB node. The SRBB VM is built from Geth enhancing its transaction management. The consensus is the Red Belly Blockchain's consensus CNG21. ➊ A client sends a transaction to some replica(s), ➋ at each replica, the transaction server receives the transactions and sends them to the transaction pool that validates transactions and ➌ sends a block to the consensus client. ➍ The consensus client $\mathord {\it propose}$s it to the consensus protocol. ➏ When the consensus outputs some acceptable blocks, ➐ all of these blocks are combined into a superblock and sent for execution. Each block in the superblock is validated and executed➑. Subsequently, the block is inserted into a chain and stored in the data store➒.
• Figure 2: Performance difference when processing each block of a superblock at a time (optimized) and when processing the entire superblock at once (non-optimized)
• Figure 3: The latency (top) and throughput (bottom) comparison between SRBB (optimized) and the default geth EVM equipped with DBFT to decide superblocks (non optimized)
• Figure 4: Scalability of SRBB as its latency (left) and throughput (right) under varying sending rates (from 15,000 TPS to 35,000 TPS) when deployed on 200 machines spread in 5 continents
• Figure 5: Comparison of modern blockchains latency (left) and throughput (right) when running the exchange decentralized application (DApp) with the Nasdaq trace

Theorems & Definitions (5)

• Definition 1: The Blockchain Problem
• Lemma 1
• Theorem 1
• Theorem 2
• Theorem 3