High-performance smart contracts concurrent execution for permissioned blockchain using SGX

Since there are no security concerns such as Sybil attacks, selfish mining, etc., the higher the system throughput, the better for the permissioned blockchain. And with the emergence of consensus algorithms, the throughput rates of permissioned blockchain can be up to thousands of transactions per second. The existing serial execution method for smart contracts becomes a new bottleneck for the system. Due to the lack of mutual trust between nodes, for a batch of smart contracts contained in a block, the traditional two-phase smart contract concurrency approach can only achieve concurrency within a single node, but not the parallel execution of contracts between nodes. In this paper, we propose a new two-phase framework based on trusted hardware Intel SGX, which can avoid the re-execution of all smart contracts on all nodes and improve parallelism between nodes. And consistency between nodes is achieved directly with state replication, rather than by re-executing transactions. We design a pre-execution mechanism for smart contracts in untrusted memory to batch fetch all the state data that a smart contract needs to access to reduce frequent enclave transitions during smart contract execution. Besides, we propose a method that generates a compact read-write set and a data structure named Merkle Forest which can generate the compact Merkle multiproofs for the initial data in untrusted memory in parallel and can quickly verify the correctness of the data passed in the enclave. Finally, we integrate all the techniques proposed in this paper into an open-source system BFT-SMaRt to evaluate our approach in a distributed setting. Experimental results show the efficiency of the proposed methods.