Lark writes Flink SQL in such a way that jobs are packaged and submitted to the Flink cluster, supporting multiple data sources and multiple storage.