CACHING TECHNIQUES USING A MAPPING CACHE AND MAINTAINING CACHE COHERENCY USING PHYSICAL TO LOGICAL ADDRESS MAPPING

    公开(公告)号:US20240303199A1

    公开(公告)日:2024-09-12

    申请号:US18119565

    申请日:2023-03-09

    IPC分类号: G06F12/0891 G06F12/0893

    摘要: Caching techniques can include: receiving a read I/O operation requesting to read from a logical address; determining whether a mapping cache includes a descriptor for the logical address, the mapping cache including two hash tables; responsive to determining that the mapping cache includes the descriptor for the logical address, performing read hit mapping cache processing which includes using a physical location identified by the descriptor to read content stored at the logical address; and responsive to determining that the mapping cache does not include the descriptor for the logical address, performing read miss mapping cache processing which includes adding the descriptor to the mapping cache. Read miss mapping cache processing can include adding the descriptor to a first hash table, used to access the descriptor by the logical address, and adding the descriptor to a second hash table used to access the descriptor by the physical location.

    VARIABLE SIZE METADATA PAGES WITH LOG-STRUCTURED METADATA

    公开(公告)号:US20240256190A1

    公开(公告)日:2024-08-01

    申请号:US18101826

    申请日:2023-01-26

    IPC分类号: G06F3/06 G06F12/1045

    摘要: In at least one embodiment, a log-structured system for metadata (MD) updates can be leveraged to enable use of variable size MD pages with multiple different encoding formats. Processing can include: generating an updated MD page by performing one or more updates to an existing version of the MD page; selecting a first page type to be used for encoding the updated MD page, wherein said selecting selects the first page type from a plurality of predefined page types, wherein each of the plurality of predefined page types identifies a different metadata layout or format for storing MD pages; encoding the updated MD page in accordance with the first page type and accordingly generating a first encoded version of the updated MD page; and storing the first encoded version of the updated MD page at a first physical storage location of a non-volatile storage device.

    GARBAGE COLLECTION PROCESSING IN STORAGE SYSTEMS

    公开(公告)号:US20240126687A1

    公开(公告)日:2024-04-18

    申请号:US17964136

    申请日:2022-10-12

    IPC分类号: G06F12/02 G06F3/06

    摘要: An apparatus comprises a processing device configured to initiate garbage collection for data pages stored in local storage of a storage node of a storage system. The processing device is also configured to determine, for a given data page stored in the local storage of the storage node, a validity score characterizing a size of changed data in the given data page, and to compare the validity score for the given data page to at least one designated threshold. The processing device is further configured to update a given page object for the given data page in an object store of persistent storage responsive to a first comparison result, and to generate, in the object store of the persistent storage, a page delta object for the given data page responsive to a second comparison result, the page delta object comprising the changed data in the given data page.

    FLEXIBLE RAID SCHEME ALLOWING FAST REBUILD
    4.
    发明公开

    公开(公告)号:US20230350752A1

    公开(公告)日:2023-11-02

    申请号:US17730388

    申请日:2022-04-27

    IPC分类号: G06F3/06 G06F9/30

    摘要: A technique protects user data within a storage system. The technique involves, in response to a request signal that directs the storage system to store the user data, generating first parity and second parity based on the user data. The technique further involves storing the first parity and the user data in a data uber of the storage system, the data uber providing redundant array of independent disks (RAID) protection. The technique further involves storing the second parity in designated free storage of the storage system, the designated free storage being separate from the data uber.

    Priority based raid (redundant array of independent disks) rebuild for a data storage system

    公开(公告)号:US12073081B2

    公开(公告)日:2024-08-27

    申请号:US18070664

    申请日:2022-11-29

    IPC分类号: G06F3/06

    摘要: Non-volatile data storage of a data storage system is organized into multiple RAID extents. Each RAID extent contains multiple logically contiguous RAID stripes. In response to detecting a failed drive in the data storage system, an initial rebuilding pass is performed on each one of the RAID extents. The initial rebuilding pass recovers data that was previously stored on the failed drive and was located within high priority RAID stripes in the RAID extent. After the initial rebuilding pass is completed on all of the RAID extents, at least one subsequent rebuilding pass is performed on each one of the RAID extents. The subsequent rebuilding pass recovers data that was previously stored on the failed drive and was located within RAID stripes in the RAID extent that are not high priority.

    Scaling midplane bandwidth between storage processors via network devices

    公开(公告)号:US12072827B2

    公开(公告)日:2024-08-27

    申请号:US17971795

    申请日:2022-10-24

    IPC分类号: G06F13/40 G06F13/42

    摘要: Techniques provide communications bandwidth between storage processors (SPs). Such techniques involve electrically coupling the SPs with a first side of a midplane. Such techniques further involve electrically coupling a network interface controller (NIC) device with a second side of the midplane that is opposite the first side of the midplane. Such techniques further involve configuring the NIC device to convey communications between the SPs while the SPs are electrically coupled with the first side of the midplane and while the NIC device is electrically coupled with the second side of the midplane that is opposite the first side of the midplane.

    Variable size metadata pages with log-structured metadata

    公开(公告)号:US12061821B1

    公开(公告)日:2024-08-13

    申请号:US18101826

    申请日:2023-01-26

    摘要: In at least one embodiment, a log-structured system for metadata (MD) updates can be leveraged to enable use of variable size MD pages with multiple different encoding formats. Processing can include: generating an updated MD page by performing one or more updates to an existing version of the MD page; selecting a first page type to be used for encoding the updated MD page, wherein said selecting selects the first page type from a plurality of predefined page types, wherein each of the plurality of predefined page types identifies a different metadata layout or format for storing MD pages; encoding the updated MD page in accordance with the first page type and accordingly generating a first encoded version of the updated MD page; and storing the first encoded version of the updated MD page at a first physical storage location of a non-volatile storage device.

    Techniques for improving write performance using zone sharing in log structured systems

    公开(公告)号:US11995329B1

    公开(公告)日:2024-05-28

    申请号:US17994794

    申请日:2022-11-28

    IPC分类号: G06F12/00 G06F3/06

    摘要: A data storage system can use non-volatile solid state drives (SSDs) to provide backend storage. The data storage system and SSDs can implement log structured systems (LSSs) experiencing write amplification (WA). The SSDs can be partitioned into zones. To reduce WA of the SSD-internal LSS, multiple nodes can write to a single active zone. The active zone can be partitioned into portions each assigned for exclusive use by one of the nodes. Each node can allocate storage from its corresponding portion of the active zone. When a node consumes a threshold amount of its portion, the active zone can be repartitioned in accordance with storage consumption, write or allocation rates of the nodes. When the active zone does not have a minimum amount of free space, the active zone can be finally repartitioned. A node can switch its active zone to a next zone once it consumes its final partition.

    Reversible write techniques using physical storage device offloading

    公开(公告)号:US11853592B2

    公开(公告)日:2023-12-26

    申请号:US17715538

    申请日:2022-04-07

    IPC分类号: G06F3/06

    摘要: A system can use non-volatile solid state drives (SSDs) to provide storage. The SSDs can implement internal log structured systems (LSSs). A reversible write operation can be serviced by an SSD to write first data to an SSD logical address. The reversible write operation can update the SSD logical address to store the first data rather than old data stored at the SSD logical address prior to servicing the reversible write operation. The old data can be stored at an SSD physical address and mapping information indicates the SSD logical address is mapped to the SSD physical address. Servicing the reversible write operation can include: retaining the mapping information and the old data; and adding second mapping information that maps the SSD logical address to a second SSD physical address storing the first data. A subsequent read operation can read the old data using the retained mapping information.

    Techniques for flexible physical drive expansion using a loop back connection

    公开(公告)号:US11726660B1

    公开(公告)日:2023-08-15

    申请号:US17721785

    申请日:2022-04-15

    IPC分类号: G06F3/06 G06F13/10

    摘要: Techniques providing connectivity between a CPU and physical storage devices (PDs) can use a loop back path formed between two connectors of an extended PO slot when an extended I/O card is inserted therein. The two connectors can include a first connector having connectivity with the CPU over a first set of lanes, and a second connector having connectivity with the PDs over a second set of lanes. While the extended I/O card is inserted into the I/O slot, connectivity can be provided between the CPU and the PDs using connectivity provided between the CPU and the first connector and the first set of lanes; using the loop back path provided between the first and second connectors; and using connectivity provided between the second connector and the PDs over the second set of lanes.