Abstract: The present invention provides a method, data processing system and computer program product in which a plurality of software executable libraries (sub-libraries) are combined into a single software executable library. The single library comprises a plurality of register functions, each associated with a particular sub-library and each sub-library comprises one or more functions. When the library receives a request to register from a library manager it calls each of the register functions, each of which register the one or more functions of its associated sub-library with the library manager. Advantageously the library manager can subsequently interface with each sub-library as if the sub-library was a standalone library. Further a register function can be used for a library whether the library is sub-library or a standalone library, thereby reducing the effort to switch from one to the other.

Abstract: A method and system for managing data in a data storage system including a plurality of physical storage devices is disclosed. The method and system comprises providing a plurality of modules, where each of the plurality of modules includes at least one child. Each module receives an input command related to the data from a source, where the source is transparent to the module. The module decides to which one of its children to pass the input command, and passes the input command to that child for processing the data accordingly.

Abstract: Communication from a processor in a computer system to a remote input/output (I/O) unit in an expansion drawer using a Peripheral Component Interface (PCI) protocol is optimized to improve system performance. An InfiniBand (IB) protocol link is used to couple the I/O unit to the computer system. In one case the computer system uses a PCI to IB bridge to couple from the processor to a corresponding IB to PCI bridge in the expansion drawer which couples to the I/O unit using a PCI link. Intelligence is added to the PCI to IB bridge to optimize communication by assigning selected PCI command sequences to optimized Macro IB commands. The IB to PCI link has like intelligence to receive and convert the Macro IB commands to the corresponding selected PCI command sequences. Optimization is accomplished by either a learning routine or by a logic state machine that generate the optimized Macro commands.

Abstract: A method and system for balancing deferred procedure queues in multiprocessor computer systems provides a greater use of multiprocessing power for the handling of hardware requests in a device driver. By creating a number of DPC (deferred procedure call) objects and work lists greater than or equal to the number of processors, a device driver can queue work items to the work lists in a round-robin fashion and maintain a queued DPC object for each work list containing work items. This balances processor assignment to DPC execution, providing enough work lists and associated DPC objects to the device driver so that each processor may concurrently perform DPC tasks for the device, or so that the DPC tasks are balanced with other system or application tasks each processor is assigned.

Abstract: A method and system for replacing a device driver during system operation provides for hot-swapping device drivers without shutting down a device driver stack. Particularly useful with file system and file storage devices, the method and system utilize a shell device driver that replaces the physical device driver's entry point interface. The shell device driver has a device driver side interface that calls the physical device driver in response to operating system requests. When an update of the physical device driver is requested, either by an operating system utility or a special upgrade application, the shell device driver receives an update indication. The shell device driver can then finish required tasks, uncouple and unload the existing physical device driver and couple to the new device driver without disrupting the operating system device interface.

Abstract: A method and system for managing data in a data storage system including a plurality of physical storage devices is disclosed. The method and system comprises providing a plurality of modules, where each of the plurality of modules includes at least one child. Each module receives an input command related to the data from a source, where the source is transparent to the module. The module decides to which one of its children to pass the input command, and passes the input command to that child for processing the data accordingly.