Design And Implementation Of A High Availability Distrubuted File System

In a distributed file system, servers may crash for various reasons. So it is a key point to distributed file system that how can the cluster provide service to outside normally when internal errors occur.Based on the research of existing distributed file systems, this thesis proposes a high available distributed file system based on a two-layer index structure. With a series of strategies, this distributed file system is capable of handling serveral failure cases, such as network connection failure, network turbulence, and cluster server's denial of service, which promotes system availability by design, the main work of this thesis is as follow:1. Secondary index structure is proposed to reduce the load of Master node in conventional distributed file system, which improves system availability indirectly.2. Different roles are being ensured the in different unusual circumstances, such as an internal network jitters or network failures, single cluster or multiple physical servers crashes. support mechanism to quickly complete the transfer failure, so the system can continue to provide service to clients, which is ensured high system availability.3. Multiple copies of the files and load balancing mechanism is being used in system at the same time to increase availability throughout the distributed file system.The tests demonstrate that this system can recover promptly from downtime of one role or multiple roles, hence ensuring the whole system high availability.