Abstract: A computer system having a scaleable software architecture is disclosed. The scaleable communication or data replication architecture that enables transparent replication of data or state information over a network of geographically dispersed processing units. Transparent data replication over a geographically dispersed computer network is useful in applications such as parallel computing and disaster recovery. The communication architecture also provides a transparent interface to a kernel I/O subsystem, device drivers and system applications. The communication architecture provides a distributed data model presenting a single system image of the I/O subsystem that allows two or more geographically dispersed processing units or clusters thereof, access to common data. In one particular implementation, the communication architecture permits RAID algorithms, such as RAID level 1 and RAID level 5 state information to be applied to the geographically dispersed network for site disaster recovery.

Abstract: Upon a computer system crash, the operating system writes data such as stored in an in-core disk to an external nonvolatile disk drive. During a reboot of the system, volatile memory is cleared; and the location of the stored user data in the nonvolatile disk drive is determined. The location of the in-core disk in the nonvolatile disk drive is identified and the in-core disk's data is written back to memory, thereby preventing the loss of the user data due to a system crash.