Abstract: In order to manage massive storage efficiently, Blue Whale distributed file system discards the traditional central resource management model, and adopts a distributed layered resource management model. This model supports multiple storage nodes and a cluster of metadata servers. The out-of-band data transportation alleviates the bottlenecks of performance, and enables metadata server cluster to handle metadata concurrently and efficiently, and also provides load balancing in the system. Theoretical analysis and test results show that this model outperforms in capability and scalability in various circumstances.