Chain of Trust: Can Trusted Hardware Help Scaling Blockchains?
share
As blockchain systems proliferate, there remains an unresolved scalability problem of their underlying distributed consensus protocols. Byzantine Fault Tolerance (BFT) consensus protocols offer high transaction throughput, but can only support small networks. Proof-of-Work (PoW) consensus protocol, on the other hand, can support thousands of nodes, but at the expense of poor transaction throughput. Two potential approaches to address these scalability barriers are by relaxing the threat model, or employing sharding technique to deal with large networks. Nonetheless, their effectiveness against data-intensive blockchain workloads remains to be seen. In this work, we study the use and effectiveness of trusted hardware on scaling distributed consensus protocols, and by their extension, blockchain systems. We first analyze existing approaches that harness trusted hardware to enhances scalability, and identify their limitations. Drawing insights from these results, we propose two design principles, namely scale up by scaling down and prioritize consensus messages, that enable the consensus protocols to scale. We illustrate the two principles by presenting optimizations that improve upon state-of-the-art solutions, and demonstrate via our extensive evaluations that they indeed offer better scalability. In particular, integrating our optimizations into Hyperledger Fabric achieves up to 7x higher throughput, while enabling the system to remain operational as the network size increases. Another optimization that we introduce to Hyperledger Sawtooth allows the system to sustain high throughput as the network grows. Finally, our new design for sharding protocols reduces the cost of shard creation phase by upto 35x.
READ FULL TEXT
