Abstract: A “Logging” method and apparatus is provided to protect control unit cached data not yet written to backing storage disk drives. This recording mechanism copies “WRITE DATA” to a log at a target logically external or physically external to storage control units which are connected to a common storage pool and constitute a set. Upon failure of one of the set of control units, the “Log” information is available to insure that pending “write” data is written to the proper location on the disk drives of the storage pool upon a recovery action. One of surviving members of the set assumes control of the storage managed by the failing control unit by utilizing the recorded information to insure that data not written to backing storage (disks) up to the point of failure is then written to the disk backing storage.

Abstract: A device and method for connection to a host and transferring data between the host and data storage assembly. The device is preferably a storage system which initializes a logical unit (LUN) in the storage assembly. A persistent memory stores a data representation of clean areas in the LUN. The storage system receives data to be written from the host to a destination area in the LUN, and determines if the destination area in the LUN is a clean area or a dirty area as indicated in the persistent memory. The storage system effects a snapshot copy of data from the destination area in the LUN to an archive storage device of the storage assembly only if the destination area is not entirely a clean area. Bandwidth is likewise reduced in LUN copy and reconstruction operations.

Abstract: A RAID system which functions despite any single point of failure is disclosed. The system has two or more controllers, a multiplicity of direct access storage devices arranged in racks, redundant connectors throughout, and an independent backplane, cables, power supply, and cooling system for each controller and for each rack of direct access storage devices. In a second embodiment, no backplane or midplane is included in the RAID system; rather each controller is connected directly with the direct access storage devices.

Abstract: The RAID system of this invention is able to function with undiminished capacity despite the failure of any one DASD, connector, or storage array controller. A signal is produced when any component fails, thereby alerting the operator of the need to replace the failed component. The reliability of this RAID system is achieved by using two connectors for each connection, by using a spare DASD for each channel of DASD, and by providing a passive storage array controller able to assume the identify of and take over the control of the DASD controlled by a failed storage array controller. This is a reliable, inexpensive system for achieving unprecedented reliability in the storage and delivery of data.

Abstract: The RAID system disclosed here uses arbitrated fiber channels or switch fabric to connect multiple host computers and storage array controllers (SAC). Each SAC is designated a primary SAC for an array of storage units, which it normally serves as controller, and as a secondary SAC for another array of storage units. A primary SAC, secondary SAC, and array of storage units is termed a storage unit set. When the primary SAC or associated host computer fails, the failure is detected by an interface chip, which causes the secondary SAC to assume the identify of the primary controller. Using system configuration information from the DASDs, the secondary SAC then controls the storage units of the storage unit set along with the storage units of which it is primary SAC. With this configuration, there is no need for switch apparatus between the storage arrays and there is no interference because dual ported storage units are used.

Abstract: In order to provide redundancy of host computers and storage array controllers, and thereby protect operation of a RAID system against host computer or storage array controller failure, two single RAID subsystems are conventionally provided in a conventional active-active configuration. This means each storage array controller has electrical access to each disk in its array as well as each disk in the other controller's array. The resultant interference between the controllers in accessing the disk channels causes the active-active system to normally function at approximately 130% of the speed of a single RAID system, rather than the optimum 200% of the speed of a single RAID system. The system of this invention, FULL-SPEED ACTIVE-ACTIVE redundant RAID system, contains a normally open switch or repeater which allows access by each storage array controller only to that controller's array of disks when both host computers and storage array controllers are operating normally.

Abstract: This invention is a software-based method for facilitating the recovery of a RAID storage system from the simultaneous failure of two or more disks (catastrophic failure). It involves the identification of the logical address and length of the failed areas of the failed disks and the writing of this information into a bad region table which is replicated on each disk. This makes it easier and less expensive to identify the problem areas and make the necessary repairs.

Abstract: This invention provides a software-based method for determining the identity of a failed (hanging) channel in an array of multiple, buffered, synchronized data channels. A hanging channel may stem from a failed storage unit in the channel, a faulty connection within the channel, or a fault in the interface circuitry which connects the channel with the synchronization hardware. It is necessary and useful to identify a hanging channel in order to isolate it (as in a RAID 2-5 system) or for repair. The hanging channel is detected by examination of the byte content of the buffers. If the system is in a read mode, the buffer serving the hanging channel will contain substantially more bytes than the other buffers. If the system is in the write mode, the buffer serving the hanging channel will contain substantially fewer bytes than the other buffers.