R. Y. Wang, T. E. Anderson. xFS: A Wide Area Mass Storage File System. Proc. Fourth Workshop on Workstation Operating Systems pp. 71-78. October 1993. Also appeared as University of California Technical Report CSD-93-783.
The current generation of file systems are inadequate in facing the new technological challenges of wide area networks and massive storage. WANs have lower bandwidth, higher latency, and higher cost. Such a scarce resource must be used intelligently. The management of a large number of hosts and massive amount of data demands a system with good scalability. And as the system is made more scalable, allowing larger groups of clients and servers to work together, it also becomes more likely at any given time that some hosts will be unable to communicate. Existing disk-based local area network file systems fail to distinguish the difference between WAN communication and LAN traffic and generate excessive traffic over the WAN links. They depend on central server models that offer poor scalability and availability. Although some systems do offer limited tertiary support, such features are usually based on ad hoc extensions to disk-based data structures and the result is usually less than satisfactory. We are currently implementing xFS, a prototype wide area mass storage file system. Our goal in xFS is to provide the performance and availability of a local disk file system when sharing is minimum and storage requirement is small. It adapts many of the techniques used in the field of high performance multiprocessor design. xFS organizes hosts into a hierarchical structure so locality within clusters of workstations can be better exploited. By using an invalidation-based write back cache coherence protocol, xFS minimizes network usage. It exploits the file system naming structure to reduce cache coherence state. xFS integrates different storage technologies in a uniform manner. Data is located by using a translation mechanism similar to that of virtual-to-physical mapping of memory management. Storage is managed in a log-structured manner for fast write performance. Due to its intelligent use of local hosts and local storage, we expect xFS to achieve better performance and availability than current generation network file systems run in the wide area.