| 摘要 |
Multiple computers have access to a shared storage system. The computers negotiate with each other for access rights (e.g., READ or WRITE) to regions of the storage system. After negotiating for permission to modify a location (e.g., a range of addresses) of the shared storage system, a given computer copies data from the shared storage system into its own corresponding cache. The given computer granted the permission then modifies the data in its cache while maintaining a separate journal of corresponding changes to metadata. The given computer writes the journal to a dedicated region of the storage system. Thereafter, the given computer attempts to flush contents of the modified cache to the storage system. If the given computer crashes while attempting to flush the modified cache to the shared storage system, a healthy one of the multiple computers utilizes the journal information in the storage system to recover metadata.
|
