Disk shadowing coupled with the Bear Computer Systems Virtual Disk product provides remarkable capabilities. It adds all of the load-balancing and redundant-write protection of disk shadowing to the already outstanding features of its companion product, Memory Disk. The flexibility of virtual disks makes it possible to shadow only critical data rather than the whole volume. Thus, catchup operations are speeded up, less processor time and disk space is used, and there is seldom any need for additional hardware.

Disk shadowing allows disk backup to proceed during prime time without interruption of user programs. A special feature flushes file system data to the disk when the operator wishes to remove a shadow unit from the set. In conjunction with the improved performance that shadowing provides, this feature allows for a nearly resource-free method of disk backup.

Disk shadowing, when combined with the included memory disk feature, provides a true write-through cache product. Frequently accessed areas of the disk are available at memory-transfer rates while the data are not vulnerable to the volatility of memory. This product provide s the most complete set of tools available for controlling disk access, backup, performance and reliability -- no matter what the size or hardware mix of the system!

Cluster-wide disk shadowing adds yet another invaluable facet to the MagicDisk product line-up. It allows disk shadowing to run on each n ode of a cluster, and each node can further have a locally-resident copy of selected shadow sets. An additional process to send and receive distributed lock requests permits shadow sets to be as available as clustered hardware.

Cluster-wide disk shadowing adds these exciting additional benefits to disk shadowing: multiple nodes can service I/O operations to shadow sets; catchup operations need only run on one processor; and different processors may be selected to simultaneously do catchup operations for different shadow sets.

Cluster-wide disk shadowing also makes operator intervention unnecessary to discontinue or re-establish shadowing operations when a node l eaves or enters the cluster. The status of shadow sets is kept consistent on all nodes, and provided there is a physical path to at least one shadow member, the shadow set continues on all nodes.