Availability in the Echo File System

The Echo file system project explored several issues in the design and implementation of distributed file systems. This paper describes the aspects of the Echo design that are related to providing high availability. These aspects include the provision of redundant components (replicated disks and backup servers), the replication of information, and recovery from failures. Further, we discuss some less obvious mechanisms needed for providing truly high availability: load control, dynamic reconfiguration of the system, and the detection and reporting of faults. Finally, we discuss some of the impact of our availability mechanisms on application software.