Abstract: Method and apparatus for asynchronous discovery of processing and storage nodes coupled via an expander switch in a fabric. In some embodiments, an initiator device operates as a processing node to transfer data to and from a non-volatile memory (NVM) of a target device at a storage node. One of the initiator or target devices is activated prior to the other device. The second activated device broadcasts a discovery command responsive to the activation of the second activated device and prior to receipt of a request for the discovery command from the first activated device. The first activated device processes the discovery command to establish an I/O communication link between the first activated device and the second activated device. The discovery command may include a non-volatile memory express (NVMe) controller list, and the NVM may be arranged as one or more NVMe namespaces.

Abstract: Methods, systems, and computer-readable storage media for performing scattered atomic I/O writes in a storage device. A list of block I/O write requests to be completed as an atomic unit is received from a requester with at least two of the block I/O write requests specifying non-contiguous data locations on a storage media. The plurality of block I/O write requests are buffered in a write buffer with each buffer entry marked as having an invalid state, wherein marking a buffer entry as having an invalid state prevents it from being flushed to the storage media. Upon buffering all of the plurality of block I/O writes, all of the buffer entries are marked as having a valid state at the same time. Upon marking all of the buffer entries as having a valid state, successful completion of the list of block I/O write requests is acknowledged to the requester.

Abstract: Method and apparatus for asynchronous discovery of processing and storage nodes coupled via an expander switch in a fabric. In some embodiments, an initiator device operates as a processing node to transfer data to and from a non-volatile memory (NVM) of a target device at a storage node. One of the initiator or target devices is activated prior to the other device. The second activated device broadcasts a discovery command responsive to the activation of the second activated device and prior to receipt of a request for the discovery command from the first activated device. The first activated device processes the discovery command to establish an I/O communication link between the first activated device and the second activated device. The discovery command may include a non-volatile memory express (NVMe) controller list, and the NVM may be arranged as one or more NVMe namespaces.

Abstract: Methods, systems, and computer-readable storage media for performing scattered atomic I/O writes in a storage device. A list of block I/O write requests to be completed as an atomic unit is received from a requester with at least two of the block I/O write requests specifying non-contiguous data locations on a storage media. The plurality of block I/O write requests are buffered in a write buffer with each buffer entry marked as having an invalid state, wherein marking a buffer entry as having an invalid state prevents it from being flushed to the storage media. Upon buffering all of the plurality of block I/O writes, all of the buffer entries are marked as having a valid state at the same time. Upon marking all of the buffer entries as having a valid state, successful completion of the list of block I/O write requests is acknowledged to the requester.

Abstract: The disclosure is directed to a system and method of cache management for a data storage system. According to various embodiments, the cache management system includes a hinting driver, a priority controller, and a data scrubber. The hinting driver generates pointers based upon data packets intercepted from data transfer requests being processed by a host controller of the data storage system. The priority controller determines whether the data transfer request includes a request to discard a portion of data based upon the pointers generated by the hinting driver. If the priority controller determines that data transfer request includes a request to discard a portion of data, the data scrubber locates and removes the portion of data from the cache memory so that the cache memory is freed from invalid data (e.g. data associated with a deleted file).

Abstract: Methods and structure for splitting Input/Output (I/O) for Redundant Array of Independent Disks (RAID) systems. One embodiment is a system that includes a processor of a host, and a memory of the host. The processor and the memory implement a device driver for communicating with a physically distinct RAID controller. The device driver is able to receive an I/O request, from an Operating System of the host, that is directed to a RAID volume. The device driver is further able to determine that the controller includes dedicated circuitry for handling I/O requests directed to a single RAID strip. Responsive to determining that the controller includes such dedicated circuitry, the device driver is able to identify RAID strip boundaries within the received request, and to generate multiple child I/O requests that are each directed to a single strip of the volume and correspond to the identified strip boundaries.

Abstract: The disclosure is directed to a system and method of cache management for a data storage system. According to various embodiments, the cache management system includes a hinting driver, a priority controller, and a data scrubber. The hinting driver generates pointers based upon data packets intercepted from data transfer requests being processed by a host controller of the data storage system. The priority controller determines whether the data transfer request includes a request to discard a portion of data based upon the pointers generated by the hinting driver. If the priority controller determines that data transfer request includes a request to discard a portion of data, the data scrubber locates and removes the portion of data from the cache memory so that the cache memory is freed from invalid data (e.g. data associated with a deleted file).