Hyperledger++ is a permissioned blockchain platform. Based on Hyperledger v0.6, it is optimized for performance at scale. In particular, Hyperledger++ features a scalable Byzantine consensus protocol that delivers up to 7x better throughput than the original Hyperledger. Furthermore, it can scale up to tolerate many more failures. The platform uses trusted hardware, i.e. Intel SGX, to lower the trust requirement for the consensus protocol, thereby reducing network communication and enabling higher scalability.
Current blockchain system is hamstrung by the performance of the underlying Byzantine consensus protocol, as recently observed in BLOCKBENCH. Classical Byzantine Fault Tolerance protocol, namely PBFT, is communication bound, which means it fails to scale to more than a handful of nodes. In particular, to tolerate f failures, the protocol needs 3f+1 nodes to communicate. Hyperledger v0.6 implements PBFT, therefore suffering the same limitation.
Hyperledger++ improves PBFT performance by using trusted hardware to reduce the required network size to 2f+1. As a consequence, for a given f Hyperledger++ incurs less communication overhead than the original Hyperledger. More importantly, it allows the platform to scale to higher f, which means it can tolerate more failures.
There are two key design choices behind Hyperledger++ performance and scalability:
- Trusted hardware based consensus protocol: Hyperledger++ implements A2M protocol that reduces the required network size from 3f+1 to 2f+1.
- Prioritize consensus over other messages: Hyperledger++ guarantees that consensus messages have higher priority, i.e. guaranteed to be processed quickly, over other network request messages.
Please check our white paper for more details.