Western Digital Technologies, Inc.

Jerry Lo

Lan Dinh Phan

Subhash Balakrishna Pillai

Yun-Tzuo Lai

Dominic S Suryabudi

Kai-Lung Cheng

Eugene Lisitsyn

Johnny A Lam

Ho-Fan Kang

Frederick H Adi

Richard David Barndt

Haining Liu

Aldo Giovanni Cometti

Lyndon S Chiu

Hung-min Chang

Hung-Cheng Yeh

Jing Booth

Parvaneh Alavi

Cliff Pajaro

Erick Omar Abasto

Stephen P Hack

A data storage subsystem is disclosed that implements a process for storing and/or reconstructing system data, such as a system mapping table. In certain embodiments, table pages are systematically copied, or flushed, to non-volatile memory in a progressive manner, according to a fixed ratio of flushed table pages per table update trigger, thereby facilitating write and/or load efficiency. Full or partial reconstruction of a table may be performed within a bounded number of operations based on the size of the table, the ratio implemented, and/or other characteristics.

System data storage mechanism providing coherency and segmented data loading

The subject technology provides for managing a data storage system. A host write command to write host data associated with a logical address to a non-volatile memory is received. A first physical address in the non-volatile memory mapped to the logical address in an address mapping table is determined. An indicator that the first physical address is bad checked. If the first physical address is indicated as bad, a valid count associated with a first set of physical addresses at a current value is maintained. The first set of physical addresses comprises the first physical address. If the first physical address is not indicated as bad, the first physical address is marked as invalid. The valid count associated with the first set of physical addresses is decremented.

Managing data for a data storage system

Host data segments are received and stored in a cached data unit corresponding to a previously stored data unit currently stored in non-volatile memory. Metadata is created that identifies unmodified previously stored segments of host data in the previously stored data unit that correspond to the received host data segments, the metadata including an update flag indicating that the previously stored data unit requires updating. In response to detecting the unexpected interruption of power, the cached data unit and the metadata is written to an area of the non-volatile memory array that is different than where the previously stored data unit is currently stored. Upon resuming operation following the unexpected interruption of power, the cached data unit is identified based on the update flag, as having been saved in response to the power shutdown without the previously stored data unit being updated in the non-volatile memory array, and then reloaded into the memory cache.

Preserving data upon a power shutdown

Aspects of the disclosure provide a method and an apparatus that perform a background media scan (BGMS) with improved efficiency. In particular, the disclosed BGMS processes can monitor data retention performance of a large capacity solid state drive (SSD) without significantly increasing scanning overhead by scanning only some sample pages of a memory block.

Method and apparatus for monitoring non-volatile memory read errors using background media scan

Managing system data for a data storage system

Aspects of the disclosure provide methods and apparatus for handling Read Disturb and block errors in a non-volatile memory (NVM) device. An error level of both an aggressor page that causes Read Disturb errors and an error level of adjacent victim pages are obtained. The error level of the victim page is compared against a predetermined threshold error level to determine if the victim page is experiencing a high level of bit errors. If so, then the error level of the aggressor page is compared to the error level of the victim page to determine whether Read Disturb errors are actually occurring due to host reads of the aggressor page. By looking at both the aggressor and victim error levels, a more accurate determination of Read Disturb errors may be obtained, resulting in less unnecessary relocations of pages and blocks within an NVM for mitigating Read Disturb effects.

Methods and apparatus for read disturb detection and handling

Aspects of the disclosure provide methods and apparatus that monitor and mitigate Read Disturb errors in non-volatile memory (NVM) devices such as NAND flash memories. The disclosed methods and apparatus determine which logical block addresses (LBAs) in the NVM device are frequently accessed by a host, rather than looking a physical address accesses. The potential Read Disturb errors may then be mitigated by triggering Read Disturb mitigation when the numbers of access of one or more of the frequently accessed LBAs exceeds a predefined number of accesses.

Methods and apparatus for read disturb detection based on logical domain

A data storage system is disclosed that utilizes garbage collection and logs for managing system data. In one embodiment, system data stored in a non-volatile memory is updated based on the character of changes to data stored in a data storage system (e.g., changes caused by host system activity). For example, when changes to stored data are scattered (e.g., changes are made to random memory locations), it may be beneficial to generate and accumulate more logs reflecting changes to the system data. As another example, when changes to stored data are substantially consolidated (e.g., changes are made to consecutive memory locations), it may be beneficial to update system data stored in the non-volatile memory more frequently. Reduction in write amplification, increase in efficiency, and reduction in start-up and initialization time can be attained. Reconstruction time of system data can also be reduced.

System data management using garbage collection and logs

An example device includes a memory device and one or more processors. The memory device is configured to store a table that includes two or more mappings, each mapping being associated with a respective logical address and a respective physical address. The processors are configured to identify, within the table, a first zone and a second zone. Each zone includes one or more mappings of the table. The zones do not share any mapping of the table. The processors are further configured to form a first log list indicating one or more mapping updates associated with the mapping(s) included in the first zone, to form a second log list indicating one or more mapping updates associated with the mapping(s) included in the second zone, and to replay a portion of the first log list and a portion of the second log list concurrently to update the table.

Zoning of logical to physical data address translation tables with parallelized log list replay

Embodiments of the invention are directed to optimizing the selection of memory blocks for garbage collection in solid state devices to efficiently maximize the amount of memory freed by garbage collection operations. The systems and methods disclosed herein provide for the efficient selection of optimal or near-optimal garbage collection candidate blocks, with the most optimal selection defined as block(s) with the most invalid pages. In one embodiment, a controller classifies memory blocks into various invalid block pools by the amount of invalid pages each block contains. In one embodiment, the controller selects for garbage collection a block from a non-empty pool of blocks with the highest minimum amount of invalid pages. One or more of the pools have minimum thresholds that can be dynamically adjusted according to an observed usage condition, such as a change of an over-provisioning amount in the storage state device.

System and method for dynamically adjusting garbage collection policies in solid-state memory

Embodiments of the invention are directed to optimizing reconstruction of operation data in volatile memory of solid-state storage subsystems. In various embodiments, operation data is stored in the volatile memory with persistent backup data of the operation data in the non-volatile memory. In one embodiment, operation data includes a superblock table that is used to identify most or all groups of blocks (superblocks) within the storage device that certain firmware components operate on. Sometimes operation data in the volatile memory is lost or corrupted due to a power interruption or system shutdown. To optimize the reconstruction of the superblock table or other similar operation data in the volatile memory, embodiments of the invention use a 'snapshot entry' to identify the latest entry information, allowing the controller to quickly identify the most updated physical locations of the operation data portions and complete the reconstruction in an efficient manner.

Growing community of inventors

Hacienda Heights, CA, United States of America

Jerry Lo