Blockchain technology has the characteristics of decentralization, security and non-modifiable. The consensus algorithm is the core of blockchain framework, which solves the problem of mutual trust among nodes in the distributed network. Selecting the appropriate consensus algorithm or not directly affects the performance of blockchain. However, existing private blockchain platform lacks theory and data support for the performance analysis of the consensus algorithm. This paper proposes a method to evaluate the performance of consensus algorithm in private blockchain platforms of Ethereum and Hyperledger Fabric. Through quantitative analysis of latency and throughput, we obtain the performance evaluation results of consensus algorithms with different numbers of transactions. The results show that the consensus mechanism induces performance bottleneck. Besides, Practical Byzantine Fault Tolerance (PBFT) consistently outperforms Proof-of-Work (PoW) in terms of latency and throughput under varying workloads. This paper contributes to providing not only the quantitative data support for the further research of consensus algorithm, but also the guidance to the blockchain practitioners in the selection of consensus mechanism.