Abstract
With the development of high-speed networks, people begin to use networks to combine all storage devices to form a huge distributed file system. Now distributed network file systems become more and more important and influential. Besides, in recent years, the concept of object-based storage has been brought up. It emphasizes all things in disks should be seen as objects, and file systems can manage all file data via these objects ignoring the storage details. Now many people believe object-based storage will become a main trend in the next generation of hard drives. With this, people begin to combine distributed network file systems with the concept of object-based storage. Now, more and more object-based distributed network file systems have been presented. As time goes by, most distributed file systems become more and more scalable, reliable and high-performance. But until now, there are still no file systems supporting fast arbitrary data insertion and truncation which reduce the time for file modification significantly. So we present Wofs, an object-based distributed network file system with support for fast arbitrary data insertion and truncation. Wofs splits a file into many small objects, stores these objects in remote file servers, and uses a special B+tree to manage the metadata of these objects. With this design, Wofs can provide good and stable performance for data insertion and truncation. Besides, Wofs uses the object-range locking policy to avoid data incoherence and improve the performance. At the final part of this thesis, the performance evaluations show Wofs has stable performance and high throughput on data insertion and truncation, which implies that Wofs is efficient and neat.