Aiming at the problem of low efficiency of application log operation and maintenance in cloud platform at present, a log management scheme based on ELKB architecture was studied. The scheme can realize the efficient collection and unified storage of log data, and provide the functions of log query and visual analysis. The log collection and storage scheme was described in detail, and the method of optimizing the basic parameters of the collector Filebeat was proposed. The log stored procedure deploys the Elasticsearch cluster mode, and summarizes the performance improvement method of the Elasticsearch cluster node. Based on this, a log visualization management system was built. The experimental results show that the memory occupancy rate is reduced by 47% compared with the default configuration when the Filebeat with optimized parameters is used for log collection. The data throughput of Elasticsearch cluster mode in keyword query (250 req/s) is higher than that of single server, and the segment merging optimization of cluster nodes effectively reduces the index memory occupancy rate of logs, and the memory optimization percentage is between 11.9% and 22.1%. It can be seen that the deployment of log management system is flexible, which effectively improves the efficiency of log collection and retrieval, and provides a feasible solution for log management of cloud platform.